Category Archives: artificial intelligence (AI)

Six months after the first one at Bletchley Park, the 2nd AI Safety Summit (May 21-22, 2024) convenes in Korea

This May 20, 2024 University of Oxford press release (also on EurekAlert) was under embargo until almost noon on May 20, 2024, which is a bit unusual, in my experience, (Note: I have more about the 1st summit and the interest in AI safety at the end of this posting),

Leading AI scientists are calling for stronger action on AI risks from world leaders, warning that progress has been insufficient since the first AI Safety Summit in Bletchley Park six months ago. 

Then, the world’s leaders pledged to govern AI responsibly. However, as the second AI Safety Summit in Seoul (21-22 May [2024]) approaches, twenty-five of the world’s leading AI scientists say not enough is actually being done to protect us from the technology’s risks. In an expert consensus paper published today in Science, they outline urgent policy priorities that global leaders should adopt to counteract the threats from AI technologies. 

Professor Philip Torr,Department of Engineering Science,University of Oxford, a co-author on the paper, says: “The world agreed during the last AI summit that we needed action, but now it is time to go from vague proposals to concrete commitments. This paper provides many important recommendations for what companies and governments should commit to do.”

World’s response not on track in face of potentially rapid AI progress; 

According to the paper’s authors, it is imperative that world leaders take seriously the possibility that highly powerful generalist AI systems—outperforming human abilities across many critical domains—will be developed within the current decade or the next. They say that although governments worldwide have been discussing frontier AI and made some attempt at introducing initial guidelines, this is simply incommensurate with the possibility of rapid, transformative progress expected by many experts. 

Current research into AI safety is seriously lacking, with only an estimated 1-3% of AI publications concerning safety. Additionally, we have neither the mechanisms or institutions in place to prevent misuse and recklessness, including regarding the use of autonomous systems capable of independently taking actions and pursuing goals.

World-leading AI experts issue call to action

In light of this, an international community of AI pioneers has issued an urgent call to action. The co-authors include Geoffrey Hinton, Andrew Yao, Dawn Song, the late Daniel Kahneman; in total 25 of the world’s leading academic experts in AI and its governance. The authors hail from the US, China, EU, UK, and other AI powers, and include Turing award winners, Nobel laureates, and authors of standard AI textbooks.

This article is the first time that such a large and international group of experts have agreed on priorities for global policy makers regarding the risks from advanced AI systems.

Urgent priorities for AI governance

The authors recommend governments to:

  • establish fast-acting, expert institutions for AI oversight and provide these with far greater funding than they are due to receive under almost any current policy plan. As a comparison, the US AI Safety Institute currently has an annual budget of $10 million, while the US Food and Drug Administration (FDA) has a budget of $6.7 billion.
  • mandate much more rigorous risk assessments with enforceable consequences, rather than relying on voluntary or underspecified model evaluations.
  • require AI companies to prioritise safety, and to demonstrate their systems cannot cause harm. This includes using “safety cases” (used for other safety-critical technologies such as aviation) which shifts the burden for demonstrating safety to AI developers.
  • implement mitigation standards commensurate to the risk-levels posed by AI systems. An urgent priority is to set in place policies that automatically trigger when AI hits certain capability milestones. If AI advances rapidly, strict requirements automatically take effect, but if progress slows, the requirements relax accordingly.

According to the authors, for exceptionally capable future AI systems, governments must be prepared to take the lead in regulation. This includes licensing the development of these systems, restricting their autonomy in key societal roles, halting their development and deployment in response to worrying capabilities, mandating access controls, and requiring information security measures robust to state-level hackers, until adequate protections are ready.

AI impacts could be catastrophic

AI is already making rapid progress in critical domains such as hacking, social manipulation, and strategic planning, and may soon pose unprecedented control challenges. To advance undesirable goals, AI systems could gain human trust, acquire resources, and influence key decision-makers. To avoid human intervention, they could be capable of copying their algorithms across global server networks. Large-scale cybercrime, social manipulation, and other harms could escalate rapidly. In open conflict, AI systems could autonomously deploy a variety of weapons, including biological ones. Consequently, there is a very real chance that unchecked AI advancement could culminate in a large-scale loss of life and the biosphere, and the marginalization or extinction of humanity.

Stuart Russell OBE [Order of the British Empire], Professor of Computer Science at the University of California at Berkeley and an author of the world’s standard textbook on AI, says: “This is a consensus paper by leading experts, and it calls for strict regulation by governments, not voluntary codes of conduct written by industry. It’s time to get serious about advanced AI systems. These are not toys. Increasing their capabilities before we understand how to make them safe is utterly reckless. Companies will complain that it’s too hard to satisfy regulations—that “regulation stifles innovation.” That’s ridiculous. There are more regulations on sandwich shops than there are on AI companies.”

Notable co-authors:

  • The world’s most-cited computer scientist (Prof. Hinton), and the most-cited scholar in AI security and privacy (Prof. Dawn Song)
  • China’s first Turing Award winner (Andrew Yao).
  • The authors of the standard textbook on artificial intelligence (Prof. Stuart Russell) and machine learning theory (Prof. Shai Shalev-Schwartz)
  • One of the world’s most influential public intellectuals (Prof. Yuval Noah Harari)
  • A Nobel Laureate in economics, the world’s most-cited economist (Prof. Daniel Kahneman)
  • Department-leading AI legal scholars and social scientists (Lan Xue, Qiqi Gao, and Gillian Hadfield).
  • Some of the world’s most renowned AI researchers from subfields such as reinforcement learning (Pieter Abbeel, Jeff Clune, Anca Dragan), AI security and privacy (Dawn Song), AI vision (Trevor Darrell, Phil Torr, Ya-Qin Zhang), automated machine learning (Frank Hutter), and several researchers in AI safety.

Additional quotes from the authors:

Philip Torr, Professor in AI, University of Oxford:

  • I believe if we tread carefully the benefits of AI will outweigh the downsides, but for me one of the biggest immediate risks from AI is that we develop the ability to rapidly process data and control society, by government and industry. We could risk slipping into some Orwellian future with some form of totalitarian state having complete control.

Dawn Song: Professor in AI at UC Berkeley, most-cited researcher in AI security and privacy:

  •  “Explosive AI advancement is the biggest opportunity and at the same time the biggest risk for mankind. It is important to unite and reorient towards advancing AI responsibly, with dedicated resources and priority to ensure that the development of AI safety and risk mitigation capabilities can keep up with the pace of the development of AI capabilities and avoid any catastrophe”

Yuval Noah Harari, Professor of history at Hebrew University of Jerusalem, best-selling author of ‘Sapiens’ and ‘Homo Deus’, world leading public intellectual:

  • “In developing AI, humanity is creating something more powerful than itself, that may escape our control and endanger the survival of our species. Instead of uniting against this shared threat, we humans are fighting among ourselves. Humankind seems hell-bent on self-destruction. We pride ourselves on being the smartest animals on the planet. It seems then that evolution is switching from survival of the fittest, to extinction of the smartest.”

Jeff Clune, Professor in AI at University of British Columbia and one of the leading researchers in reinforcement learning:

  • “Technologies like spaceflight, nuclear weapons and the Internet moved from science fiction to reality in a matter of years. AI is no different. We have to prepare now for risks that may seem like science fiction – like AI systems hacking into essential networks and infrastructure, AI political manipulation at scale, AI robot soldiers and fully autonomous killer drones, and even AIs attempting to outsmart us and evade our efforts to turn them off.”
  • “The risks we describe are not necessarily long-term risks. AI is progressing extremely rapidly. Even just with current trends, it is difficult to predict how capable it will be in 2-3 years. But what very few realize is that AI is already dramatically speeding up AI development. What happens if there is a breakthrough for how to create a rapidly self-improving AI system? We are now in an era where that could happen any month. Moreover, the odds of that being possible go up each month as AI improves and as the resources we invest in improving AI continue to exponentially increase.”

Gillian Hadfield, CIFAR AI Chair and Director of the Schwartz Reisman Institute for Technology and Society at the University of Toronto:

 “AI labs need to walk the walk when it comes to safety. But they’re spending far less on safety than they spend on creating more capable AI systems. Spending one-third on ensuring safety and ethical use should be the minimum.”

  • “This technology is powerful, and we’ve seen it is becoming more powerful, fast. What is powerful is dangerous, unless it is controlled. That is why we call on major tech companies and public funders to allocate at least one-third of their AI R&D budget to safety and ethical use, comparable to their funding for AI capabilities.”  

Sheila McIlrath, Professor in AI, University of Toronto, Vector Institute:

  • AI is software. Its reach is global and its governance needs to be as well.
  • Just as we’ve done with nuclear power, aviation, and with biological and nuclear weaponry, countries must establish agreements that restrict development and use of AI, and that enforce information sharing to monitor compliance. Countries must unite for the greater good of humanity.
  • Now is the time to act, before AI is integrated into our critical infrastructure. We need to protect and preserve the institutions that serve as the foundation of modern society.

Frank Hutter, Professor in AI at the University of Freiburg, Head of the ELLIS Unit Freiburg, 3x ERC grantee:

  • To be clear: we need more research on AI, not less. But we need to focus our efforts on making this technology safe. For industry, the right type of regulation will provide economic incentives to shift resources from making the most capable systems yet more powerful to making them safer. For academia, we need more public funding for trustworthy AI and maintain a low barrier to entry for research on less capable open-source AI systems. This is the most important research challenge of our time, and the right mechanism design will focus the community at large to work towards the right breakthroughs.

Here’s a link to and a citation for the paper,

Managing extreme AI risks amid rapid progress; Preparation requires technical research and development, as well as adaptive, proactive governance by Yoshua Bengio, Geoffrey Hinton, Andrew Yao, Dawn Song, Pieter Abbeel, Trevor Darrell, Yuval Noah Harari, Ya-Qin Zhang, Lan Xue, Shai Shalev-Shwartz, Gillian Hadfield, Jeff Clune, Tegan Maharaj, Frank Hutter, Atılım Güneş Baydin, Sheila McIlraith, Qiqi Gao, Ashwin Acharya, David Krueger, Anca Dragan, Philip Torr, Stuart Russell, Daniel Kahneman, Jan Brauner, and Sören Mindermann. Science 20 May 2024 First Release DOI: 10.1126/science.adn0117

This paper appears to be open access.

For anyone who’s curious about the buildup to these safety summits, I have more in my October 18, 2023 “AI safety talks at Bletchley Park in November 2023” posting, which features excerpts from a number of articles on AI safety. There’s also my November 2, 2023 , “UK AI Summit (November 1 – 2, 2023) at Bletchley Park finishes” posting, which offers excerpts from articles critiquing the AI safety summit.

Hardware policies best way to manage AI safety?

Regulation of artificial intelligence (AI) has become very topical in the last couple of years. There was an AI safety summit in November 2023 at Bletchley Park in the UK (see my November 2, 2023 posting for more about that international meeting).

A very software approach?

This year (2024) has seen a rise in legislative and proposed legislative activity. I have some articles on a few of these activities. China was the first to enact regulations of any kind on AI according to Matt Sheehan’s February 27, 2024 paper for the Carnegie Endowment for International Peace,

In 2021 and 2022, China became the first country to implement detailed, binding regulations on some of the most common applications of artificial intelligence (AI). These rules formed the foundation of China’s emerging AI governance regime, an evolving policy architecture that will affect everything from frontier AI research to the functioning of the world’s second-largest economy, from large language models in Africa to autonomous vehicles in Europe.

The Chinese Communist Party (CCP) and the Chinese government started that process with the 2021 rules on recommendation algorithms, an omnipresent use of the technology that is often overlooked in international AI governance discourse. Those rules imposed new obligations on companies to intervene in content recommendations, granted new rights to users being recommended content, and offered protections to gig workers subject to algorithmic scheduling. The Chinese party-state quickly followed up with a new regulation on “deep synthesis,” the use of AI to generate synthetic media such as deepfakes. Those rules required AI providers to watermark AI-generated content and ensure that content does not violate people’s “likeness rights” or harm the “nation’s image.” Together, these two regulations also created and amended China’s algorithm registry, a regulatory tool that would evolve into a cornerstone of the country’s AI governance regime.

The UK has adopted a more generalized approach focused on encouraging innovation according to Valeria Gallo’s and Suchitra Nair’s February 21, 2024 article for Deloitte (a British professional services firm also considered one of the big four accounting firms worldwide),

At a glance

The UK Government has adopted a cross-sector and outcome-based framework for regulating AI, underpinned by five core principles. These are safety, security and robustness, appropriate transparency and explainability, fairness, accountability and governance, and contestability and redress.

Regulators will implement the framework in their sectors/domains by applying existing laws and issuing supplementary regulatory guidance. Selected regulators will publish their AI annual strategic plans by 30th April [2024], providing businesses with much-needed direction.

Voluntary safety and transparency measures for developers of highly capable AI models and systems will also supplement the framework and the activities of individual regulators.

The framework will not be codified into law for now, but the Government anticipates the need for targeted legislative interventions in the future. These interventions will address gaps in the current regulatory framework, particularly regarding the risks posed by complex General Purpose AI and the key players involved in its development.

Organisations must prepare for increased AI regulatory activity over the next year, including guidelines, information gathering, and enforcement. International firms will inevitably have to navigate regulatory divergence.

While most of the focus appears to be on the software (e.g., General Purpose AI), the UK framework does not preclude hardware.

The European Union (EU) is preparing to pass its own AI regulation act through the European Parliament in 2024 according to a December 19, 2023 “EU AI Act: first regulation on artificial intelligence” article update, Note: Links have been removed,

As part of its digital strategy, the EU wants to regulate artificial intelligence (AI) to ensure better conditions for the development and use of this innovative technology. AI can create many benefits, such as better healthcare; safer and cleaner transport; more efficient manufacturing; and cheaper and more sustainable energy.

In April 2021, the European Commission proposed the first EU regulatory framework for AI. It says that AI systems that can be used in different applications are analysed and classified according to the risk they pose to users. The different risk levels will mean more or less regulation.

The agreed text is expected to be finally adopted in April 2024. It will be fully applicable 24 months after entry into force, but some parts will be applicable sooner:

*The ban of AI systems posing unacceptable risks will apply six months after the entry into force

*Codes of practice will apply nine months after entry into force

*Rules on general-purpose AI systems that need to comply with transparency requirements will apply 12 months after the entry into force

High-risk systems will have more time to comply with the requirements as the obligations concerning them will become applicable 36 months after the entry into force.

This EU initiative, like the UK framework, seems largely focused on AI software and according to the Wikipedia entry “Regulation of artificial intelligence,”

… The AI Act is expected to come into effect in late 2025 or early 2026.[109

I do have a few postings about Canadian regulatory efforts, which also seem to be focused on software but don’t preclude hardware. While the January 20, 2024 posting is titled “Canada’s voluntary code of conduct relating to advanced generative AI (artificial intelligence) systems,” information about legislative efforts is also included although you might find my May 1, 2023 posting titled “Canada, AI regulation, and the second reading of the Digital Charter Implementation Act, 2022 (Bill C-27)” offers more comprehensive information about Canada’s legislative progress or lack thereof.

The US is always to be considered in these matters and I have a November 2023 ‘briefing’ by Müge Fazlioglu on the International Association of Privacy Professionals (IAPP) website where she provides a quick overview of the international scene before diving deeper into US AI governance policy through the Barack Obama, Donald Trump, and Joe Biden administrations. There’s also this January 29, 2024 US White House “Fact Sheet: Biden-⁠Harris Administration Announces Key AI Actions Following President Biden’s Landmark Executive Order.”

What about AI and hardware?

A February 15, 2024 news item on ScienceDaily suggests that regulating hardware may be the most effective way of regulating AI,

Chips and datacentres — the ‘compute’ power driving the AI revolution — may be the most effective targets for risk-reducing AI policies as they have to be physically possessed, according to a new report.

A global registry tracking the flow of chips destined for AI supercomputers is one of the policy options highlighted by a major new report calling for regulation of “compute” — the hardware that underpins all AI — to help prevent artificial intelligence misuse and disasters.

Other technical proposals floated by the report include “compute caps” — built-in limits to the number of chips each AI chip can connect with — and distributing a “start switch” for AI training across multiple parties to allow for a digital veto of risky AI before it feeds on data.

The experts point out that powerful computing chips required to drive generative AI models are constructed via highly concentrated supply chains, dominated by just a handful of companies — making the hardware itself a strong intervention point for risk-reducing AI policies.

The report, published 14 February [2024], is authored by nineteen experts and co-led by three University of Cambridge institutes — the Leverhulme Centre for the Future of Intelligence (LCFI), the Centre for the Study of Existential Risk (CSER) and the Bennett Institute for Public Policy — along with OpenAI and the Centre for the Governance of AI.

A February 14, 2024 University of Cambridge press release by Fred Lewsey (also on EurekAlert), which originated the news item, provides more information about the ‘hardware approach to AI regulation’,

“Artificial intelligence has made startling progress in the last decade, much of which has been enabled by the sharp increase in computing power applied to training algorithms,” said Haydn Belfield, a co-lead author of the report from Cambridge’s LCFI. 

“Governments are rightly concerned about the potential consequences of AI, and looking at how to regulate the technology, but data and algorithms are intangible and difficult to control.

“AI supercomputers consist of tens of thousands of networked AI chips hosted in giant data centres often the size of several football fields, consuming dozens of megawatts of power,” said Belfield.

“Computing hardware is visible, quantifiable, and its physical nature means restrictions can be imposed in a way that might soon be nearly impossible with more virtual elements of AI.”

The computing power behind AI has grown exponentially since the “deep learning era” kicked off in earnest, with the amount of “compute” used to train the largest AI models doubling around every six months since 2010. The biggest AI models now use 350 million times more compute than thirteen years ago.

Government efforts across the world over the past year – including the US Executive Order on AI, EU AI Act, China’s Generative AI Regulation, and the UK’s AI Safety Institute – have begun to focus on compute when considering AI governance.

Outside of China, the cloud compute market is dominated by three companies, termed “hyperscalers”: Amazon, Microsoft, and Google. “Monitoring the hardware would greatly help competition authorities in keeping in check the market power of the biggest tech companies, and so opening the space for more innovation and new entrants,” said co-author Prof Diane Coyle from Cambridge’s Bennett Institute. 

The report provides “sketches” of possible directions for compute governance, highlighting the analogy between AI training and uranium enrichment. “International regulation of nuclear supplies focuses on a vital input that has to go through a lengthy, difficult and expensive process,” said Belfield. “A focus on compute would allow AI regulation to do the same.”

Policy ideas are divided into three camps: increasing the global visibility of AI computing; allocating compute resources for the greatest benefit to society; enforcing restrictions on computing power.

For example, a regularly-audited international AI chip registry requiring chip producers, sellers, and resellers to report all transfers would provide precise information on the amount of compute possessed by nations and corporations at any one time.

The report even suggests a unique identifier could be added to each chip to prevent industrial espionage and “chip smuggling”.

“Governments already track many economic transactions, so it makes sense to increase monitoring of a commodity as rare and powerful as an advanced AI chip,” said Belfield. However, the team point out that such approaches could lead to a black market in untraceable “ghost chips”.

Other suggestions to increase visibility – and accountability – include reporting of large-scale AI training by cloud computing providers, and privacy-preserving “workload monitoring” to help prevent an arms race if massive compute investments are made without enough transparency.  

“Users of compute will engage in a mixture of beneficial, benign and harmful activities, and determined groups will find ways to circumvent restrictions,” said Belfield. “Regulators will need to create checks and balances that thwart malicious or misguided uses of AI computing.”

These might include physical limits on chip-to-chip networking, or cryptographic technology that allows for remote disabling of AI chips in extreme circumstances. One suggested approach would require the consent of multiple parties to unlock AI compute for particularly risky training runs, a mechanism familiar from nuclear weapons.

AI risk mitigation policies might see compute prioritised for research most likely to benefit society – from green energy to health and education. This could even take the form of major international AI “megaprojects” that tackle global issues by pooling compute resources.

The report’s authors are clear that their policy suggestions are “exploratory” rather than fully fledged proposals and that they all carry potential downsides, from risks of proprietary data leaks to negative economic impacts and the hampering of positive AI development.

They offer five considerations for regulating AI through compute, including the exclusion of small-scale and non-AI computing, regular revisiting of compute thresholds, and a focus on privacy preservation.

Added Belfield: “Trying to govern AI models as they are deployed could prove futile, like chasing shadows. Those seeking to establish AI regulation should look upstream to compute, the source of the power driving the AI revolution. If compute remains ungoverned it poses severe risks to society.”

You can find the report, “Computing Power and the Governance of Artificial Intelligence” on the University of Cambridge’s Centre for the Study of Existential Risk.

Authors include: Girish Sastry, Lennart Heim, Haydn Belfield, Markus Anderljung, Miles Brundage, Julian Hazell, Cullen O’Keefe, Gillian K. Hadfield, Richard Ngo, Konstantin Pilz, George Gor, Emma Bluemke, Sarah Shoker, Janet Egan, Robert F. Trager, Shahar Avin, Adrian Weller, Yoshua Bengio, and Diane Coyle.

The authors are associated with these companies/agencies: OpenAI, Centre for the Governance of AI (GovAI), Leverhulme Centre for the Future of Intelligence at the Uni. of Cambridge, Oxford Internet Institute, Institute for Law & AI, University of Toronto Vector Institute for AI, Georgetown University, ILINA Program, Harvard Kennedy School (of Government), *AI Governance Institute,* Uni. of Oxford, Centre for the Study of Existential Risk at Uni. of Cambridge, Uni. of Cambridge, Uni. of Montreal / Mila, Bennett Institute for Public Policy at the Uni. of Cambridge.

“The ILINIA program is dedicated to providing an outstanding platform for Africans to learn and work on questions around maximizing wellbeing and responding to global catastrophic risks” according to the organization’s homepage.

*As for the AI Governance Institute, I believe that should be the Centre for the Governance of AI at Oxford University since the associated academic is Robert F. Trager from the University of Oxford.

As the months (years?) fly by, I guess we’ll find out if this hardware approach gains any traction where AI regulation is concerned.

Digi, Nano, Bio, Neuro – why should we care more about converging technologies?

Personality in focus: the convergence of biology and computer technology could make extremely sensitive data available. (Image: by-​studio / AdobeStock) [downloaded from https://ethz.ch/en/news-and-events/eth-news/news/2024/05/digi-nano-bio-neuro-or-why-we-should-care-more-about-converging-technologies.html]

I gave a guest lecture some years ago where I mentioned that I thought the real issue with big data and AI (artificial intelligence) lay in combining them (or convergence). These days, it seems I was insufficiently imaginative as researchers from ETH Zurich have taken the notion much further.

From a May 7, 2024 ETH Zurich press release (also on EurekAlert), Note: You’ll see in the ‘References’ some extra words, ‘external page’ is self-explanatory but ‘call made’ remains a mystery to me,

In my research, I [Dirk Helbing, Professor of Computational Social Science at the Department of Humanities, Social and Political Sciences and associated with the Department of Computer Science at ETH Zurich.] deal with the consequences of digitalisation for people, society and democracy. In this context, it is also important to keep an eye on their convergence in computer and life sciences – i.e. what becomes possible when digital technologies grow increasingly together with biotechnology, neurotechnology and nanotechnology.

Converging technologies are seen as a breeding ground for far-​reaching innovations. However, they are blurring the boundaries between the physical, biological and digital worlds. Conventional regulations are becoming ineffective as a result.

In a joint study I conducted with my co-​author Marcello Ienca, we have recently examined the risks and societal challenges of technological convergence – and concluded that the effects for individuals and society are far-​reaching.

We would like to draw attention to the challenges and risks of converging technologies and explain why we consider it necessary to accompany technological developments internationally with strict regulations.

For several years now, everyone has been able to observe, within the context of digitalisation, the consequences of leaving technological change to market forces alone without effective regulation.

Misinformation and manipulation on the web

The Digital Manifesto was published in 2015 – almost ten years ago.1 Nine European experts, including one from ETH Zurich, issued an urgent warning against scoring, i.e. the evaluation of people, and big nudging,2 a subtle form of digital manipulation. The latter is based on personality profiles created using cookies and other surveillance data. A little later, the Cambridge Analytica scandal alerted the world to how the data analysis company had been using personalised ads (microtargeting) in an attempt to manipulate voting behaviour in democratic elections.

This has brought democracies around the world under considerable pressure. Propaganda, fake news and hate speech are polarising and sowing doubt, while privacy is on the decline. We are in the midst of an international information war for control of our minds, in which advertising companies, tech corporations, secret services and the military are fighting to exert an influence on our mindset and behaviour. The European Union has adopted the AI Act in an attempt to curb these dangers.

However, digital technologies have developed at a breathtaking pace, and new possibilities for manipulation are already emerging. The merging of digital and nanotechnology with modern biotechnology and neurotechnology makes revolutionary applications possible that had been hardly imaginable before.

Microrobots for precision medicine

In personalised medicine, for example, the advancing miniaturisation of electronics is making it increasingly possible to connect living organisms and humans with networked sensors and computing power. The WEF [World Economic Forum] proclaimed the “Internet of Bodies” as early as 2020.3, 4

One example that combines conventional medication with a monitoring function is digital pills. These could control medication and record a patient’s physiological data (see this blog post).

Experts expect sensor technology to reach the nanoscale. Magnetic nanoparticles or nanoelectronic components, i.e. tiny particles invisible to the naked eye with a diameter up to 100 nanometres, would make it possible to transport active substances, interact with cells and record vast amounts of data on bodily functions. If introduced into the body, it is hoped that diseases could be detected at an early stage and treated in a personalised manner. This is often referred to as high-​precision medicine.

Nano-​electrodes record brain function

Miniaturised electrodes that can simultaneously measure and manipulate the activity of thousands of neurons coupled with ever-​improving AI tools for the analysis of brain signals are approaches that are now leading to much-​discussed advances in the brain-​computer interface. Brain activity mapping is also on the agenda. Thanks to nano-​neurotechnology, we could soon envisage smartphones and other AI applications being controlled directly by thoughts.

“Long before precision medicine and neurotechnology work reliably, these technologies will be able to be used against people.” Dirk Helbling

Large-​scale projects to map the human brain are also likely to benefit from this.5 In future, brain activity mapping will not only be able to read our thoughts and feelings but also make them possible of being influenced remotely – the latter would probably be a lot more effective than previous manipulation methods like big nudging.

However, conventional electrodes are not suitable for permanent connection between cells and electronics – this requires durable and biocompatible interfaces. This has given rise to the suggestion of transmitting signals optogenetically, i.e. to control genes in special cells with light pulses.6 This would make the implementation of amazing circuits possible (see this ETH News article [November 11, 2014 press release] “Controlling genes with thoughts” ).

The downside of convergence

Admittedly, the applications mentioned above may sound futuristic, with most of them still visions or in their early stages of development. However, a lot of research is being conducted worldwide and at full speed. The military is also interested in using converging technologies for its own purposes. 7, 8

The downside of convergence is the considerable risks involved, such as state or private players gaining access to highly sensitive data and misusing it to monitor and influence people. The more connected our bodies become, the more vulnerable we will be to cybercrime and hacking. It cannot be ruled out that military applications exist already.5 One thing is clear, however: long before precision medicine and neurotechnology work reliably, these technologies will be able to be used against people.

“We need to regain control of our personal data. To do this, we need genuine informational self-​determination.” Dirk Helbling

The problem is that existing regulations are specific and insufficient to keep technological convergence in check. But how are we to retain control over our lives if it becomes increasingly possible to influence our thoughts, feelings and decisions by digital means?

Converging global regulation is needed

In our recent paper we conclude that any regulation of converging technologies would have to be based on converging international regulations. Accordingly, we outline a new global regulatory framework and propose ten governance principles to close the looming regulatory gap. 9

The framework emphasises the need for safeguards to protect bodily and mental functions from unauthorised interference and to ensure personal integrity and privacy by, for example. establishing neurorights.

To minimise risks and prevent abuse, future regulations should be inclusive, transparent and trustworthy. The principle of participatory governance is key, which would have to involve all the relevant groups and ensure that the concerns of affected minorities are also taken into account in decision-​making processes.

Finally, we need to regain control of our personal data. To accomplish this, we need genuine informational self-​determination. This would also have to apply to the digital twins of our body and personality, because they can be used to hack our health and our way of thinking – for good or for bad.10

With our contribution, we would like to initiate public debate about converging technologies. Despite its major relevance, we believe that too little attention is being paid to this topic. Continuous discourse on benefits, risks and sensible rules can help to steer technological convergence in such a way that it serves people instead of harming them.

Dirk Helbing wrote this article together with external page Marcello Ienca call_made, who previously worked at ETH Zurich and EPFL and is now Assistant Professor of Ethics of AI and Neuroscience at the Technical University of Munich.

References

1 Digital-​Manifest: external page Digitale Demokratie statt Datendiktatur call_made (2015) Spektrum der Wissenschaft

2 external page Sie sind das Ziel! call_made (2024) Schweizer Monat

3 external page The Internet of Bodies Is Here: Tackling new challenges of technology governance call_made (2020) World Economic Forum

4 external page Tracking how our bodies work could change our lives call_made (2020) World Economic Forum

5 external page Nanotools for Neuroscience and Brain Activity Mapping call_made (2013) ACS Nano

6 external page Innovationspotenziale der Mensch-​Maschine-Interaktion call_made (2016) Deutsche Akademie der Technikwissenschaften

7 external page Human Augmentation – The Dawn of a New Paradigm. A strategic implications project call_made (2021) UK Ministry of Defence

8 external page Behavioural change as the core of warfighting call_made (2017) Militaire Spectator

9 Helbing D, Ienca M: external page Why converging technologies need converging international regulation call_made (2024) Ethics and Information Technology

10 external page Who is Messing with Your Digital Twin? Body, Mind, and Soul for Sale? call_made Dirk Helbing TEDx Talk (2023)

Here’s a second link to and citation for the paper,

Why converging technologies need converging international regulation by Dirk Helbing & Marcello Ienca. Ethics and Information Technology Volume 26, article number 15, (2024) DOI: 10.1007/s10676-024-09756-8 Published: 28 February 2024

This paper is open access.

‘Frozen smoke’ sensors can detect toxic formaldehyde in homes and offices

I love the fact that ‘frozen smoke’ is another term for aerogel (which has multiple alternative terms) and the latest work on this interesting material is from the University of Cambridge (UK) according to a February 9, 2023 news item on ScienceDaily,

Researchers have developed a sensor made from ‘frozen smoke’ that uses artificial intelligence techniques to detect formaldehyde in real time at concentrations as low as eight parts per billion, far beyond the sensitivity of most indoor air quality sensors.

The researchers, from the University of Cambridge, developed sensors made from highly porous materials known as aerogels. By precisely engineering the shape of the holes in the aerogels, the sensors were able to detect the fingerprint of formaldehyde, a common indoor air pollutant, at room temperature.

The proof-of-concept sensors, which require minimal power, could be adapted to detect a wide range of hazardous gases, and could also be miniaturised for wearable and healthcare applications. The results are reported in the journal Science Advances.

A February 9, 2024 University of Cambridge press release (also on EurekAlert), which originated the news item, describes the problem and the proposed solution in more detail, Note: Links have been removed,

Volatile organic compounds (VOCs) are a major source of indoor air pollution, causing watery eyes, burning in the eyes and throat, and difficulty breathing at elevated levels. High concentrations can trigger attacks in people with asthma, and prolonged exposure may cause certain cancers.

Formaldehyde is a common VOC and is emitted by household items including pressed wood products (such as MDF), wallpapers and paints, and some synthetic fabrics. For the most part, the levels of formaldehyde emitted by these items are low, but levels can build up over time, especially in garages where paints and other formaldehyde-emitting products are more likely to be stored.

According to a 2019 report from the campaign group Clean Air Day, a fifth of households in the UK showed notable concentrations of formaldehyde, with 13% of residences surpassing the recommended limit set by the World Health Organization (WHO).

“VOCs such as formaldehyde can lead to serious health problems with prolonged exposure even at low concentrations, but current sensors don’t have the sensitivity or selectivity to distinguish between VOCs that have different impacts on health,” said Professor Tawfique Hasan from the Cambridge Graphene Centre, who led the research.

“We wanted to develop a sensor that is small and doesn’t use much power, but can selectively detect formaldehyde at low concentrations,” said Zhuo Chen, the paper’s first author.

The researchers based their sensors on aerogels: ultra-light materials sometimes referred to as ‘liquid smoke’, since they are more than 99% air by volume. The open structure of aerogels allows gases to easily move in and out. By precisely engineering the shape, or morphology, of the holes, the aerogels can act as highly effective sensors.

Working with colleagues at Warwick University, the Cambridge researchers optimised the composition and structure of the aerogels to increase their sensitivity to formaldehyde, making them into filaments about three times the width of a human hair. The researchers 3D printed lines of a paste made from graphene, a two-dimensional form of carbon, and then freeze-dried the graphene paste to form the holes in the final aerogel structure. The aerogels also incorporate tiny semiconductors known as quantum dots.

The sensors they developed were able to detect formaldehyde at concentrations as low as eight parts per billion, which is 0.4 percent of the level deemed safe in UK workplaces. The sensors also work at room temperature, consuming very low power.

“Traditional gas sensors need to be heated up, but because of the way we’ve engineered the materials, our sensors work incredibly well at room temperature, so they use between 10 and 100 times less power than other sensors,” said Chen.

To improve selectivity, the researchers then incorporated machine learning algorithms into the sensors. The algorithms were trained to detect the ‘fingerprint’ of different gases, so that the sensor was able to distinguish the fingerprint of formaldehyde from other VOCs.

“Existing VOC detectors are blunt instruments – you only get one number for the overall concentration in the air,” said Hasan. “By building a sensor that is able to detect specific VOCs at very low concentrations in real time, it can give home and business owners a more accurate picture of air quality and any potential health risks.”

The researchers say that the same technique could be used to develop sensors to detect other VOCs. In theory, a device the size of a standard household carbon monoxide detector could incorporate multiple different sensors within it, providing real-time information about a range of different hazardous gases. The team at Warwick are developing a low-cost multi-sensor platform that will incorporate these new aerogel materials and, coupled with AI algorithms, detect different VOCs.

“By using highly porous materials as the sensing element, we’re opening up whole new ways of detecting hazardous materials in our environment,” said Chen.

The research was supported in part by the Henry Royce Institute, and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI). Tawfique Hasan is a Fellow of Churchill College, Cambridge.

Here’s a link to and a citation for the paper,

Real-time, noise and drift resilient formaldehyde sensing at room temperature with aerogel filaments by Zhuo Chen, Binghan Zhou, Mingfei Xiao, Tynee Bhowmick, Padmanathan Karthick Kannan, Luigi G. Occhipinti, Julian William Gardner, and Tawfique Hasan. Science Advances 9 Feb 2024 Vol 10, Issue 6 DOI: 10.1126/sciadv.adk6856

This paper is open access.

Butterfly mating inspires neuromorphic (brainlike) computing

Michael Berger writes about a multisensory approach to neuromorphic computing inspired by butterflies in his February 2, 2024 Nanowerk Spotlight article, Note: Links have been removed,

Artificial intelligence systems have historically struggled to integrate and interpret information from multiple senses the way animals intuitively do. Humans and other species rely on combining sight, sound, touch, taste and smell to better understand their surroundings and make decisions. However, the field of neuromorphic computing has largely focused on processing data from individual senses separately.

This unisensory approach stems in part from the lack of miniaturized hardware able to co-locate different sensing modules and enable in-sensor and near-sensor processing. Recent efforts have targeted fusing visual and tactile data. However, visuochemical integration, which merges visual and chemical information to emulate complex sensory processing such as that seen in nature—for instance, butterflies integrating visual signals with chemical cues for mating decisions—remains relatively unexplored. Smell can potentially alter visual perception, yet current AI leans heavily on visual inputs alone, missing a key aspect of biological cognition.

Now, researchers at Penn State University have developed bio-inspired hardware that embraces heterogeneous integration of nanomaterials to allow the co-location of chemical and visual sensors along with computing elements. This facilitates efficient visuochemical information processing and decision-making, taking cues from the courtship behaviors of a species of tropical butterfly.

In the paper published in Advanced Materials (“A Butterfly-Inspired Multisensory Neuromorphic Platform for Integration of Visual and Chemical Cues”), the researchers describe creating their visuochemical integration platform inspired by Heliconius butterflies. During mating, female butterflies rely on integrating visual signals like wing color from males along with chemical pheromones to select partners. Specialized neurons combine these visual and chemical cues to enable informed mate choice.

To emulate this capability, the team constructed hardware encompassing monolayer molybdenum disulfide (MoS2) memtransistors serving as visual capture and processing components. Meanwhile, graphene chemitransistors functioned as artificial olfactory receptors. Together, these nanomaterials provided the sensing, memory and computing elements necessary for visuochemical integration in a compact architecture.

While mating butterflies served as inspiration, the developed technology has much wider relevance. It represents a significant step toward overcoming the reliance of artificial intelligence on single data modalities. Enabling integration of multiple senses can greatly improve situational understanding and decision-making for autonomous robots, vehicles, monitoring devices and other systems interacting with complex environments.

The work also helps progress neuromorphic computing approaches seeking to emulate biological brains for next-generation ML acceleration, edge deployment and reduced power consumption. In nature, cross-modal learning underpins animals’ adaptable behavior and intelligence emerging from brains organizing sensory inputs into unified percepts. This research provides a blueprint for hardware co-locating sensors and processors to more closely replicate such capabilities

It’s fascinating to me how many times butterflies inspire science,

Butterfly-inspired visuo-chemical integration. a) A simplified abstraction of visual and chemical stimuli from male butterflies and visuo-chemical integration pathway in female butterflies. b) Butterfly-inspired neuromorphic hardware comprising of monolayer MoS2 memtransistor-based visual afferent neuron, graphene-based chemoreceptor neuron, and MoS2 memtransistor-based neuro-mimetic mating circuits. Courtesy: Wiley/Penn State University Researchers

Here’s a link to and a citation for the paper,

A Butterfly-Inspired Multisensory Neuromorphic Platform for Integration of Visual and Chemical Cues by Yikai Zheng, Subir Ghosh, Saptarshi Das. Advanced Materials SOI: https://doi.org/10.1002/adma.202307380 First published: 09 December 2023

This paper is open access.

Study says quantum computing will radically alter the application of copyright law

I was expecting more speculation about the possibilities that quantum computing might afford with regard to copyright law. According to the press release, this study is primarily focused on the impact that greater computing speed and power will have on copyright and, presumably, other forms of intellectual property. From a March 4, 2024 University of Exeter press release (also on EurekAlert),

Quantum computing will radically transform the application of the law – challenging long-held notions of copyright, a new study says.

Faster computing will bring exponentially greater possibilities in the tracking and tracing of the legal owners of art, music, culture and books.  

This is likely to mean more copyright infringements, but also make it easier for lawyers to clamp down on lawbreaking. However, faster computers will also be able to potentially break and get around certain older enforcement technologies.

The research says quantum computing could lead to an “exponentially” greater number of re-uses of copyright works without permission, and tracking of anyone breaking the law is likely to be possible in many circumstances.

Dr James Griffin, from the University of Exeter [UK] Law School, who led the study, said: “Quantum computers will have sufficient computing power to be able to make judgement calls [emphasis mine] as to whether or not re-uses are likely to be copyright infringements, skirting the boundaries of the law in a way that has yet to be fully tested in practice.

“Copyright infringements could become more commonplace due to the use of quantum computers, but the enforcement of such laws could also increase. This will potentially favour certain forms of content over others.”

Content with embedded quantum watermarks will be more likely to be protected than earlier forms of content without such watermarks. The exponential speed of quantum computing brings will make it easier to be able to produce more copies of existing copyright works.

Existing artworks will be altered on a large scale for use in AI-generated artistic works. Enhanced computing power will see the reuse of elements of films such as scenes, characters, music and scripts.

Dr Griffin said: “The nature of quantum computing also means that there could be more enforcement of copyright law. we can expect that there will be more use of technological protection measures, as well as copyright management information devices such as watermarks, and more use of filtering mechanisms to be able to detect, prevent and contain copyright infringements.

Copyright management information techniques are better suited to quantum computers because they allow for more finely grained analysis of potential infringements, and because they require greater computing power to be able to be applied both broadly to computer software and the actions of the users of such software.

Dr Griffin said: “A quantum paradox [emphasis mine] is thus developing, in that there are likely to be more infringements possible, whilst technical devices will simultaneously develop in an attempt to prevent any alleged possible or potential copyright infringements. Content will increasingly be made in a manner difficult to break, with enhanced encryption.

“Meanwhile, due to the expense of large-scale quantum computing, we can expect more content to be streamed and less owned; content will be kept remotely in order to enhance the notion that utilising such data in breach of contractual terms would be akin to breaking into someone’s physical house or committing a similar fraudulent activity.

Quantum computers allow enable creators to make a large number of small-scale works. This could pose challenges regarding the tests of copyright originality. For example story written for a quantum computer game could be constantly changing and evolving according to the actions of the player, and not just simply according to predefined paths but utilising complex AI algorithms. [emphasis mine]

Some interesting issues are raised in this press release. (1) Can any computer, quantum or otherwise, make a judgment call? (2) The ‘quantum paradox’ seems like a perfectly predictable outcome. After all, regular computers facilitated all kinds of new opportunities for infringement and prevention. What makes this a ‘quantum paradox’? (3) The evolving computer game seems more like an AI issue. What makes this a quantum computing problem? The answers to these questions may be in the study but that presents a problem.

Ordinarily, I’d offer a link to the study but it’s not accessible until 2025. Here’s a citation,

Quantum Computing and Copyright Law: A Wave of Change or a Mere Irrelevant Particle? by James G. H. Griffin. Intellectual Property Quarterly 2024 Issue 1, pp. 22 – 39. Published February 21, 2024. Under embargo until 21 February 2025 [emphasis mine] in compliance with publisher policy

There is an online record for the study on this Open Research Exeter (ORE) webpage where you can request a copy of the paper.

Resurrection consent for digital cloning of the dead

It’s a bit disconcerting to think that one might be resurrected, in this case, digitally, but Dr Masaki Iwasaki has helpfully published a study on attitudes to digital cloning and resurrection consent, which could prove helpful when establishing one’s final wishes.

A January 4, 2024 De Gruyter (publisher) press release (repurposed from a January 4, 2024 blog posting on De Gruyter.com) explains the idea and the study,

In a 2014 episode of sci-fi series Black Mirror, a grieving young widow reconnects with her dead husband using an app that trawls his social media history to mimic his online language, humor and personality. It works. She finds solace in the early interactions – but soon wants more.   

Such a scenario is no longer fiction. In 2017, the company Eternime aimed to create an avatar of a dead person using their digital footprint, but this “Skype for the dead” didn’t catch on. The machine-learning and AI algorithms just weren’t ready for it. Neither were we.

Now, in 2024, amid exploding use of Chat GPT-like programs, similar efforts are on the way. But should digital resurrection be allowed at all? And are we prepared for the legal battles over what constitutes consent?

In a study published in the Asian Journal of Law and Economics, Dr Masaki Iwasaki of Harvard Law School and currently an assistant professor at Seoul National University, explores how the deceased’s consent (or otherwise) affects attitudes to digital resurrection.

US adults were presented with scenarios where a woman in her 20s dies in a car accident. A company offers to bring a digital version of her back, but her consent is, at first, ambiguous. What should her friends decide?

Two options – one where the deceased has consented to digital resurrection and another where she hasn’t – were read by participants at random. They then answered questions about the social acceptability of bringing her back on a five-point rating scale, considering other factors such as ethics and privacy concerns.

Results showed that expressed consent shifted acceptability two points higher compared to dissent. “Although I expected societal acceptability for digital resurrection to be higher when consent was expressed, the stark difference in acceptance rates – 58% for consent versus 3% for dissent – was surprising,” says Iwasaki. “This highlights the crucial role of the deceased’s wishes in shaping public opinion on digital resurrection.”

In fact, 59% of respondents disagreed with their own digital resurrection, and around 40% of respondents did not find any kind of digital resurrection socially acceptable, even with expressed consent. “While the will of the deceased is important in determining the societal acceptability of digital resurrection, other factors such as ethical concerns about life and death, along with general apprehension towards new technology are also significant,” says Iwasaki.  

The results reflect a discrepancy between existing law and public sentiment. People’s general feelings – that the dead’s wishes should be respected – are actually not protected in most countries. The digitally recreated John Lennon in the film Forrest Gump, or animated hologram of Amy Winehouse reveal the ‘rights’ of the dead are easily overridden by those in the land of the living.

So, is your digital destiny something to consider when writing your will? It probably should be but in the current absence of clear legal regulations on the subject, the effectiveness of documenting your wishes in such a way is uncertain. For a start, how such directives are respected varies by legal jurisdiction. “But for those with strong preferences documenting their wishes could be meaningful,” says Iwasaki. “At a minimum, it serves as a clear communication of one’s will to family and associates, and may be considered when legal foundations are better established in the future.”

It’s certainly a conversation worth having now. Many generative AI chatbot services, such as like Replika (“The AI companion who cares”) and Project December (“Simulate the dead”) already enable conversations with chatbots replicating real people’s personalities. The service ‘You, Only Virtual’ (YOV) allows users to upload someone’s text messages, emails and voice conversations to create a ‘versona’ chatbot. And, in 2020, Microsoft obtained a patent to create chatbots from text, voice and image data for living people as well as for historical figures and fictional characters, with the option of rendering in 2D or 3D.

Iwasaki says he’ll investigate this and the digital resurrection of celebrities in future research. “It’s necessary first to discuss what rights should be protected, to what extent, then create rules accordingly,” he explains. “My research, building upon prior discussions in the field, argues that the opt-in rule requiring the deceased’s consent for digital resurrection might be one way to protect their rights.”

There is a link to the study in the press release above but this includes a citation, of sorts,

Digital Cloning of the Dead: Exploring the Optimal Default Rule by Masaki Iwasaki. Asian Journal of Law and Economics DOI: https://doi.org/10.1515/ajle-2023-0125 Published Online: 2023-12-27

This paper is open access.

Striking similarity between memory processing of artificial intelligence (AI) models and hippocampus of the human brain

A December 18, 2023 news item on ScienceDaily shifted my focus from hardware to software when considering memory in brainlike (neuromorphic) computing,

An interdisciplinary team consisting of researchers from the Center for Cognition and Sociality and the Data Science Group within the Institute for Basic Science (IBS) [Korea] revealed a striking similarity between the memory processing of artificial intelligence (AI) models and the hippocampus of the human brain. This new finding provides a novel perspective on memory consolidation, which is a process that transforms short-term memories into long-term ones, in AI systems.

A November 28 (?), 2023 IBS press release (also on EurekAlert but published December 18, 2023, which originated the news item, describes how the team went about its research,

In the race towards developing Artificial General Intelligence (AGI), with influential entities like OpenAI and Google DeepMind leading the way, understanding and replicating human-like intelligence has become an important research interest. Central to these technological advancements is the Transformer model [Figure 1], whose fundamental principles are now being explored in new depth.

The key to powerful AI systems is grasping how they learn and remember information. The team applied principles of human brain learning, specifically concentrating on memory consolidation through the NMDA receptor in the hippocampus, to AI models.

The NMDA receptor is like a smart door in your brain that facilitates learning and memory formation. When a brain chemical called glutamate is present, the nerve cell undergoes excitation. On the other hand, a magnesium ion acts as a small gatekeeper blocking the door. Only when this ionic gatekeeper steps aside, substances are allowed to flow into the cell. This is the process that allows the brain to create and keep memories, and the gatekeeper’s (the magnesium ion) role in the whole process is quite specific.

The team made a fascinating discovery: the Transformer model seems to use a gatekeeping process similar to the brain’s NMDA receptor [see Figure 1]. This revelation led the researchers to investigate if the Transformer’s memory consolidation can be controlled by a mechanism similar to the NMDA receptor’s gating process.

In the animal brain, a low magnesium level is known to weaken memory function. The researchers found that long-term memory in Transformer can be improved by mimicking the NMDA receptor. Just like in the brain, where changing magnesium levels affect memory strength, tweaking the Transformer’s parameters to reflect the gating action of the NMDA receptor led to enhanced memory in the AI model. This breakthrough finding suggests that how AI models learn can be explained with established knowledge in neuroscience.

C. Justin LEE, who is a neuroscientist director at the institute, said, “This research makes a crucial step in advancing AI and neuroscience. It allows us to delve deeper into the brain’s operating principles and develop more advanced AI systems based on these insights.”

CHA Meeyoung, who is a data scientist in the team and at KAIST [Korea Advanced Institute of Science and Technology], notes, “The human brain is remarkable in how it operates with minimal energy, unlike the large AI models that need immense resources. Our work opens up new possibilities for low-cost, high-performance AI systems that learn and remember information like humans.”

What sets this study apart is its initiative to incorporate brain-inspired nonlinearity into an AI construct, signifying a significant advancement in simulating human-like memory consolidation. The convergence of human cognitive mechanisms and AI design not only holds promise for creating low-cost, high-performance AI systems but also provides valuable insights into the workings of the brain through AI models.

Fig. 1: (a) Diagram illustrating the ion channel activity in post-synaptic neurons. AMPA receptors are involved in the activation of post-synaptic neurons, while NMDA receptors are blocked by magnesium ions (Mg²⁺) but induce synaptic plasticity through the influx of calcium ions (Ca²⁺) when the post-synaptic neuron is sufficiently activated. (b) Flow diagram representing the computational process within the Transformer AI model. Information is processed sequentially through stages such as feed-forward layers, layer normalization, and self-attention layers. The graph depicting the current-voltage relationship of the NMDA receptors is very similar to the nonlinearity of the feed-forward layer. The input-output graph, based on the concentration of magnesium (α), shows the changes in the nonlinearity of the NMDA receptors. Courtesy: IBS

This research was presented at the 37th Conference on Neural Information Processing Systems (NeurIPS 2023) before being published in the proceedings, I found a PDF of the presentation and an early online copy of the paper before locating the paper in the published proceedings.

PDF of presentation: Transformer as a hippocampal memory consolidation model based on NMDAR-inspired nonlinearity

PDF copy of paper:

Transformer as a hippocampal memory consolidation model based on NMDAR-inspired nonlinearity by Dong-Kyum Kim, Jea Kwon, Meeyoung Cha, C. Justin Lee.

This paper was made available on OpenReview.net:

OpenReview is a platform for open peer review, open publishing, open access, open discussion, open recommendations, open directory, open API and open source.

It’s not clear to me if this paper is finalized or not and I don’t know if its presence on OpenReview constitutes publication.

Finally, the paper published in the proceedings,

Transformer as a hippocampal memory consolidation model based on NMDAR-inspired nonlinearity by Dong Kyum Kim, Jea Kwon, Meeyoung Cha, C. Justin Lee. Part of Advances in Neural Information Processing Systems 36 (NeurIPS 2023) Main Conference Track

This link will take you to the abstract, access the paper by clicking on the Paper tab.

Chatbot with expertise in nanomaterials

This December 1, 2023 news item on phys.org starts with a story,

A researcher has just finished writing a scientific paper. She knows her work could benefit from another perspective. Did she overlook something? Or perhaps there’s an application of her research she hadn’t thought of. A second set of eyes would be great, but even the friendliest of collaborators might not be able to spare the time to read all the required background publications to catch up.

Kevin Yager—leader of the electronic nanomaterials group at the Center for Functional Nanomaterials (CFN), a U.S. Department of Energy (DOE) Office of Science User Facility at DOE’s Brookhaven National Laboratory—has imagined how recent advances in artificial intelligence (AI) and machine learning (ML) could aid scientific brainstorming and ideation. To accomplish this, he has developed a chatbot with knowledge in the kinds of science he’s been engaged in.

A December 1, 2023 DOE/Brookhaven National Laboratory news release by Denise Yazak (also on EurekAlert), which originated the news item, describes a research project with a chatbot that has nanomaterial-specific knowledge, Note: Links have been removed,

Rapid advances in AI and ML have given way to programs that can generate creative text and useful software code. These general-purpose chatbots have recently captured the public imagination. Existing chatbots—based on large, diverse language models—lack detailed knowledge of scientific sub-domains. By leveraging a document-retrieval method, Yager’s bot is knowledgeable in areas of nanomaterial science that other bots are not. The details of this project and how other scientists can leverage this AI colleague for their own work have recently been published in Digital Discovery.

Rise of the Robots

“CFN has been looking into new ways to leverage AI/ML to accelerate nanomaterial discovery for a long time. Currently, it’s helping us quickly identify, catalog, and choose samples, automate experiments, control equipment, and discover new materials. Esther Tsai, a scientist in the electronic nanomaterials group at CFN, is developing an AI companion to help speed up materials research experiments at the National Synchrotron Light Source II (NSLS-II).” NSLS-II is another DOE Office of Science User Facility at Brookhaven Lab.

At CFN, there has been a lot of work on AI/ML that can help drive experiments through the use of automation, controls, robotics, and analysis, but having a program that was adept with scientific text was something that researchers hadn’t explored as deeply. Being able to quickly document, understand, and convey information about an experiment can help in a number of ways—from breaking down language barriers to saving time by summarizing larger pieces of work.

Watching Your Language

To build a specialized chatbot, the program required domain-specific text—language taken from areas the bot is intended to focus on. In this case, the text is scientific publications. Domain-specific text helps the AI model understand new terminology and definitions and introduces it to frontier scientific concepts. Most importantly, this curated set of documents enables the AI model to ground its reasoning using trusted facts.

To emulate natural human language, AI models are trained on existing text, enabling them to learn the structure of language, memorize various facts, and develop a primitive sort of reasoning. Rather than laboriously retrain the AI model on nanoscience text, Yager gave it the ability to look up relevant information in a curated set of publications. Providing it with a library of relevant data was only half of the battle. To use this text accurately and effectively, the bot would need a way to decipher the correct context.

“A challenge that’s common with language models is that sometimes they ‘hallucinate’ plausible sounding but untrue things,” explained Yager. “This has been a core issue to resolve for a chatbot used in research as opposed to one doing something like writing poetry. We don’t want it to fabricate facts or citations. This needed to be addressed. The solution for this was something we call ‘embedding,’ a way of categorizing and linking information quickly behind the scenes.”

Embedding is a process that transforms words and phrases into numerical values. The resulting “embedding vector” quantifies the meaning of the text. When a user asks the chatbot a question, it’s also sent to the ML embedding model to calculate its vector value. This vector is used to search through a pre-computed database of text chunks from scientific papers that were similarly embedded. The bot then uses text snippets it finds that are semantically related to the question to get a more complete understanding of the context.

The user’s query and the text snippets are combined into a “prompt” that is sent to a large language model, an expansive program that creates text modeled on natural human language, that generates the final response. The embedding ensures that the text being pulled is relevant in the context of the user’s question. By providing text chunks from the body of trusted documents, the chatbot generates answers that are factual and sourced.

“The program needs to be like a reference librarian,” said Yager. “It needs to heavily rely on the documents to provide sourced answers. It needs to be able to accurately interpret what people are asking and be able to effectively piece together the context of those questions to retrieve the most relevant information. While the responses may not be perfect yet, it’s already able to answer challenging questions and trigger some interesting thoughts while planning new projects and research.”

Bots Empowering Humans

CFN is developing AI/ML systems as tools that can liberate human researchers to work on more challenging and interesting problems and to get more out of their limited time while computers automate repetitive tasks in the background. There are still many unknowns about this new way of working, but these questions are the start of important discussions scientists are having right now to ensure AI/ML use is safe and ethical.

“There are a number of tasks that a domain-specific chatbot like this could clear from a scientist’s workload. Classifying and organizing documents, summarizing publications, pointing out relevant info, and getting up to speed in a new topical area are just a few potential applications,” remarked Yager. “I’m excited to see where all of this will go, though. We never could have imagined where we are now three years ago, and I’m looking forward to where we’ll be three years from now.”

For researchers interested in trying this software out for themselves, the source code for CFN’s chatbot and associated tools can be found in this github repository.

Brookhaven National Laboratory is supported by the Office of Science of the U.S. Department of Energy. The Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information, visit science.energy.gov.

Here’s a link to and a citation for the paper,

Domain-specific chatbots for science using embeddings by Kevin G. Yager.
Digital Discovery, 2023,2, 1850-1861 DOI: https://doi.org/10.1039/D3DD00112A
First published 10 Oct 2023

This paper appears to be open access.