Sunday Sweep: Week in Review (March 31, 2024)

Ankesh Chandaria

March 31, 2024

(

April 20, 2024

tl;dr

With this week's condensed version of the Sunday Sweep, we focus on a couple of key developments. First, Neuralink's released an incredible video showing its first human patient, Noland Arbaugh, manipulating a cursor on a monitor entirely with his mind. This sort of technology demands serious consideration both for the tremendous benefits it can bring as well as the potential ethical issues that come hand-in-hand with it. On the regulatory side, representatives of the US Congress proposed the Protecting Consumers from Deceptive AI Act -- federal legislation designed to protect the public from deepfakes. The White House Office of Management and Budget issued its first major AI policy, and a pair of additional government departments released their own AI reports (including a Treasury report on managing AI cybersecurity risk in the financial services sector).

This Sunday I'm bringing you a shorter version of the sweep (back to the regular schedule. soon!) focusing on a couple major stories from the past few weeks.

Neuralink Update: A Glimpse of The Future

The standout news for me from the last couple of weeks remains Neuralink. Back in January, Neuralink announced that its first human patient — Noland Arbaugh, a 29-year-old paralyzed from the shoulders down — had received an implant and was on the road to recovery [1]. On March 20th, Neuralink went a step further, sharing a captivating live update video of Noland. In it, he describes his accident, recovery, and what having the chip in his head has been like… all while playing a game of chess with his mind.

This is incredible stuff. But don’t take my word for it. I highly suggest you watch the video.

‍

So how does it work?

Basically, the device reads electrical patterns in the brain and associates them with actions outside. Implanted directly into the patient’s skull, it consists over 60 polymer threads which provide over a thousand sites for recording neuronal electrical signals (neural electrophysiological recording, in case it comes up on a trivia night). These patterns of neural activation are measured against the patient’s baseline and translated into various actions. Like I said. Incredible stuff.

There is, reason to move forward with caution: this technology presents a new frontier beyond one we barely understand (i.e., our brains). Indeed, we just don’t know whether and how this sort of procedure might inadvertently interfere with a patient’s brain. And what of potential misinterpretation of the signals sent through the device? A little further afield, there are concerns around hacking — could a bad actor somehow hijack and damage the device and, as a result, the human as well? I will also admit to having some discomfort at the idea of a private company controlling the contents of a device like this. Could they shut down access? Push buggy updates? A list of concerns might go on for pages...

However, it is impossible to deny the good that could come from this sort of research and the hope it offers to so many people. By fully capitalizing on machine learning technology, Neuralink demonstrates just the first step into what will likely a be a brave new world of treatment for those with paralysis, and even perhaps eventually other conditions like epilepsy and Parkinson’s. We just have to make sure we walk into that world with our eyes wide open, and the best interests of the human patients in mind.

‍

U.S. (Federal!) Government Making Moves

Protecting Consumers from Deceptive AI Act

On 21 March 2024, Congresswoman Anna Eshoo (Co-Chair of the House AI Caucus) and Congressman Neal Dunn introduced the Protecting Consumers from Deceptive AI Act. This is the first proposed federal legislation from the House (similar legislation was proposed by the Senate in September 2023 [2]) that would offer the sorts of protections around deepfakes that States have been pushing ahead with. The bill appears to be endorsed by Hugging Face, IEEE, the Center for Countering Digital Hate, and others [3].

It would, amongst other things:

Direct NIST (the National Institutes of Standards and Technology) to facilitate the development of standards for labelling and identifying AI content;
Require that generative-AI devs include machine readable disclosures within audio and visual content generated by their AI apps; and
Require online platforms to use those disclosures to label AI-generated content.

Government-Wide AI Policy

Just a week later, VP Harris announced that the White House Office of Management and Budget ("OMB") was issuing its first government-wide AI policy [4]. The policy governs AI procurement across Federal Government agencies including through "specific minimum risk management practices for uses of AI that impact the rights and safety of the public". Agencies must, for example, designate a Chief AI Officer (an individual responsible for the coordination, promotion, and risk management of the agency's use of AI) within 60 days of the issuance of the memo. A significant portion of the memo (Section 4) is dedicated to advancing responsible AI innovation with a view to removing unhelpful barriers whilst still maintaining effective guardrails -- certainly the struggle of anyone looking to regulate AI. Further, Section 5 of the memo spells out a risk mitigation approach for "safety-impacting" AI (i.e., that could impact human safety, climate, critical infrastructure) and "rights-impacting" AI (i.e., that impacts one's civil rights and liberties, privacy, access to equal opportunities, or access to critical government services).

Reports Galore

In addition to the OMB policy, the NTIA (National Telecoms and Information Administration under the Department of Commerce) published an Accountability Policy Report and the Department of Treasury set out a report on managing AI cybersecurity risks in the financial services sector.

‍

References

[1] https://www.cbc.ca/news/business/musk-brain-implant-first-human-1.7099093

[2] https://www.klobuchar.senate.gov/public/index.cfm/news-releases?ID=AF782E4C-C2C9-4C7C-8696-374F72C03F90

[3] https://eshoo.house.gov/media/press-releases/rep-eshoo-introduces-bipartisan-bill-label-deepfakes

[4] https://www.whitehouse.gov/briefing-room/statements-releases/2024/03/28/fact-sheet-vice-president-harris-announces-omb-policy-to-advance-governance-innovation-and-risk-management-in-federal-agencies-use-of-artificial-intelligence/

You may also be interested in...

A Deep and Digital Breath: Designing Responsible Anthropomorphic AI

November 1, 2024

When ChatGPT started "breathing" during voice interactions this year, this design choice seemed innocuous. However, it highlights a concern around increasingly human-like AI features being implemented without proper evaluation of their necessity or impact. The tragic case of Character.AI -- where anthropomorphic design choices allegedly contributed to a teenager's death -- demonstrates the urgent need for systematic analysis of these features. This article proposes a two-step framework for evaluating anthropomorphic AI design: first assessing both social and technical necessity, then implementing appropriate guidelines and safeguards (some of which are suggested herein). While some applications (like accessibility tools) may well justify human-like features, others (like companionship AI) tread on ethically difficult territory. As AI becomes more sophisticated, we need clear criteria for determining when anthropomorphic design serves legitimate social purposes. An LLM doesn't really need to breathe, and perhaps we ought to think much harder about how we would like human-to-human relationships to look like in the future before delving too far into enabling human-to-AI facsimiles.

Mindreading Models: Do LLMs Have Theory of Mind?

June 2, 2024

Did you know we're all mind readers? Humans develop what’s called “theory of mind” (ToM) around the age of four, enabling us to hold views and beliefs about our own mental states and the mental states of others. It’s one of those things that, well, makes us human. However, a recent paper by Strachan et al. explores this capability in the context of AI -- a finding which has some interesting implications. The authors compared the performance of a series of LLMs (GPT-3, GPT-4, and LLaMA2) against around 1,900 humans across a battery of tests designed to explore ToM. They determined that, in every task bar one (designed to test faux pas), LLMs can perform at or above human level. My own cursory testing suggests that GPT-4o may even be able to demonstrate ToM across the board. While we can’t conclude that these models actually have ToM (in a cognitive sense) they certainly seem to behave as if they do. So what does this mean in practice? These LLMs open the door to exciting potential applications across areas such as mental health and decision support making. Ultimately, this paper’s conclusions give us a moment to stop and think about the state of our technology: to be impressed by what we’ve accomplished, and acknowledge the risks that should be considered as we move forward.

Whistleblower Woes: The Right to Warn

June 5, 2024

On 4 June 2024, employees at OpenAI and Google DeepMind released an open letter imploring their companies to facilitate what they call a right to warn about advanced AI. In the absence of proper government oversight, they call upon their companies to: (1) not enter or enforce any non-disparagement clauses when it comes to risk-related concerns; (2) develop internal, verifiably anonymous, whistleblower processes; (3) support a culture of open criticism which enables disclosure of risk-related concerns; and (4) not retaliate against employees (past or present) who choose to publicly voice risk-related concerns where other processes have failed. In the uncomfortable light of commercial reality, most of these requests appear unrealistic. A demand for robust internal whistleblowing processes, however, is an actionable and arguably best next step -- particularly given the poor state of protection under the current whistleblowing regime in England and Wales. There should be a right to warn about advanced AI. This demand may well be the starting point.

Who’s Responsible for AI Safety?

May 22, 2024

The past few weeks have brought the issue of AI Safety to the fore in two contrasting ways: first, OpenAI launched GPT-4o, a promising new model that will be free for public use. Around the same time, Ilya Sutskever (OpenAI's renowned co-founder and chief scientist) left the company amidst reports that OpenAI had dissolved its long-term AI Safety team. Although the company reiterated its commitment to safe development, questions linger. Second, the UK's Department of Science, Innovation and Technology released its interim International Scientific Report on the Safety of Advanced AI. The report focuses on general-purpose AI and explores various risks as well as some technical risk mitigation techniques. It also leaves plenty unanswered. With so much uncertainty, it's imperative we seek to resolve the most practically-relevant question as soon as possible: who's responsible for AI Safety?

Sunday Sweep: Week in Review (March 17, 2024)

March 15, 2024

This week, the European Parliament passed the Artificial Intelligence Act (AIA) -- sweeping legislation governing the deployment of A.I. models and systems in the European market. The AIA also applies to A.I. companies situated outside of the E.U. that deploy their services in a way that is accessible within the E.U. Some of the provisions will kick in as soon as 6 months from when the Act enters into force. India, meanwhile, pulled back from last week's statement suggesting a turn away from onerous regulatory requirements. Electoral integrity is front of mind around the world this year, and the E.U. Commission has sent information requests to a number of major tech companies (including Google, X and TikTok) for details on generative A.I. and deepfake risks. In the U.S., Biden set out his 2025 budget request, which includes $20 billion across the board for A.I. research agencies. Also in America, a Florida man was barred from practicing law for a year for submitting hallucinated case citations and apparently being entirely unrepentant. Are we surprised? Finally, Mercedes is trialing humanoid robots for mundane tasks in a slow shuffle towards a robotic future.