Microsoft Scraps Entire Ethical AI Team Amid AI Boom

Lauren Leffer

gizmodo.com

Updated on March 14, 2023

Still relevant

Microsoft is currently in the process of shoehorning text-generating artificial intelligence into every single product that it can. And starting this month, the company will be continuing on its AI rampage without a team dedicated to internally ensuring those AI features meet Microsoft’s ethical standards, according to a Monday night report from Platformer.

Microsoft has scrapped its whole Ethics and Society team within the company’s AI sector, as part of ongoing layoffs set to impact 10,000 total employees, per Platformer. The company maintains its Office of Responsible AI, which creates the broad, Microsoft-wide principles to govern corporate AI decision making. But the ethics and society taskforce, which bridged the gap between policy and products, is reportedly no more.

Gizmodo reached out to Microsoft to confirm the news. In response, a company spokesperson sent the following statement:

Microsoft remains committed to developing and designing AI products and experiences safely and responsibly. As the technology has evolved and strengthened, so has our investment, which at times has meant adjusting team structures to be more effective. For example, over the past six years we have increased the number of people within our product teams who are dedicated to ensuring we adhere to our AI principles. We have also increased the scale and scope of our Office of Responsible AI, which provides cross-company support for things like reviewing sensitive use cases and advocating for policies that protect customers.

To Platformer, the company reportedly previously shared this slightly different version of the same statement:

Microsoft is committed to developing AI products and experiences safely and responsibly...Over the past six years we have increased the number of people across our product teams within the Office of Responsible AI who, along with all of us at Microsoft, are accountable for ensuring we put our AI principles into practice...We appreciate the trailblazing work the ethics and society team did to help us on our ongoing responsible AI journey.

Note that, in this older version, Microsoft does inadvertently confirm that the ethics and society team is no more. The company also previously specified staffing increases were in the Office of Responsible AI vs people generally “dedicated to ensuring we adhere to our AI principles.”

Yet, despite Microsoft’s reassurances, former employees told Platformer that the Ethics and Society team played a key role translating big ideas from the responsibility office into actionable changes at the product development level.

The info is here.

The Darwinian Argument for Worrying About AI

Dan Hendrycks

Time.com

Originally posted 31 May 23

Here is an excerpt:

In the biological realm, evolution is a slow process. For humans, it takes nine months to create the next generation and around 20 years of schooling and parenting to produce fully functional adults. But scientists have observed meaningful evolutionary changes in species with rapid reproduction rates, like fruit flies, in fewer than 10 generations. Unconstrained by biology, AIs could adapt—and therefore evolve—even faster than fruit flies do.

There are three reasons this should worry us. The first is that selection effects make AIs difficult to control. Whereas AI researchers once spoke of “designing” AIs, they now speak of “steering” them. And even our ability to steer is slipping out of our grasp as we let AIs teach themselves and increasingly act in ways that even their creators do not fully understand. In advanced artificial neural networks, we understand the inputs that go into the system, but the output emerges from a “black box” with a decision-making process largely indecipherable to humans.

Second, evolution tends to produce selfish behavior. Amoral competition among AIs may select for undesirable traits. AIs that successfully gain influence and provide economic value will predominate, replacing AIs that act in a more narrow and constrained manner, even if this comes at the cost of lowering guardrails and safety measures. As an example, most businesses follow laws, but in situations where stealing trade secrets or deceiving regulators is highly lucrative and difficult to detect, a business that engages in such selfish behavior will most likely outperform its more principled competitors.

Selfishness doesn’t require malice or even sentience. When an AI automates a task and leaves a human jobless, this is selfish behavior without any intent. If competitive pressures continue to drive AI development, we shouldn’t be surprised if they act selfishly too.

The third reason is that evolutionary pressure will likely ingrain AIs with behaviors that promote self-preservation. Skeptics of AI risks often ask, “Couldn’t we just turn the AI off?” There are a variety of practical challenges here. The AI could be under the control of a different nation or a bad actor. Or AIs could be integrated into vital infrastructure, like power grids or the internet. When embedded into these critical systems, the cost of disabling them may prove too high for us to accept since we would become dependent on them. AIs could become embedded in our world in ways that we can’t easily reverse. But natural selection poses a more fundamental barrier: we will select against AIs that are easy to turn off, and we will come to depend on AIs that we are less likely to turn off.

These strong economic and strategic pressures to adopt the systems that are most effective mean that humans are incentivized to cede more and more power to AI systems that cannot be reliably controlled, putting us on a pathway toward being supplanted as the earth’s dominant species. There are no easy, surefire solutions to our predicament.

The info is here.

Ethical Accident Algorithms for Autonomous Vehicles and the Trolley Problem: Three Philosophical Disputes

Sven Nyholm

In Lillehammer, H. (ed.), The Trolley Problem.

Cambridge: Cambridge University Press, 2023

Abstract

The Trolley Problem is one of the most intensively discussed and controversial puzzles in contemporary moral philosophy. Over the last half-century, it has also become something of a cultural phenomenon, having been the subject of scientific experiments, online polls, television programs, computer games, and several popular books. This volume offers newly written chapters on a range of topics including the formulation of the Trolley Problem and its standard variations; the evaluation of different forms of moral theory; the neuroscience and social psychology of moral behavior; and the application of thought experiments to moral dilemmas in real life. The chapters are written by leading experts on moral theory, applied philosophy, neuroscience, and social psychology, and include several authors who have set the terms of the ongoing debates. The volume will be valuable for students and scholars working on any aspect of the Trolley Problem and its intellectual significance.

Here is the conclusion:

Accordingly, it seems to me that just as the first methodological approach mentioned a few paragraphs above is problematic, so is the third methodological approach. In other words, we do best to take the second approach. We should neither rely too heavily (or indeed exclusively) on the comparison between the ethics of self-driving cars and the trolley problem, nor wholly ignore and pay no attention to the comparison between the ethics of self-driving cars and the trolley problem. Rather, we do best to make this one – but not the only – thing we do when we think about the ethics of self-driving cars. With what is still a relatively new issue for philosophical ethics to work with, and indeed also regarding older ethical issues that have been around much longer, using a mixed and pluralistic method that approaches the moral issues we are considering from many different angles is surely the best way to go. In this instance, that includes reflecting on – and reflecting critically on – how the ethics of crashes involving self-driving cars is both similar to and different from the philosophy of the trolley problem.

At this point, somebody might say, “what if I am somebody who really dislikes the self-driving cars/trolley problem comparison, and I would really prefer reflecting on the ethics of self-driving cars without spending any time on thinking about the similarities and differences between the ethics of self-driving cars and the trolley problem?” In other words, should everyone working on the ethics of self-driving cars spend at least some of their time reflecting on the comparison with the trolley problem? Luckily for those who are reluctant to spend any of their time reflecting on the self-driving cars/trolley problem comparison, there are others who are willing and able to devote at least some of their energies to this comparison.

In general, I think we should view the community that works on the ethics of this issue as being one in which there can be a division of labor, whereby different members of this field can partly focus on different things, and thereby together cover all of the different aspects that are relevant and important to investigate regarding the ethics of self-driving cars.  As it happens, there has been a remarkable variety in the methods and approaches people have used to address the ethics of self-driving cars (see Nyholm 2018 a-b).  So, while it is my own view that anybody who wants to form a complete overview of the ethics of self-driving cars should, among other things, devote some of their time to studying the comparison with the trolley problem, it is ultimately no big problem if not everyone wishes to do so. There are others who have been studying, and who will most likely continue to reflect on, this comparison.

A pdf is here.

Generative AI entails a credit–blame asymmetry

Porsdam Mann, S., Earp, B. et al. (2023).

Nature Machine Intelligence.

https://doi.org/10.1038/s42256-023-00653-1

The recent releases of large-scale language models (LLMs), including OpenAI’s ChatGPT and GPT-4, Meta’s LLaMA, and Google’s Bard have garnered substantial global attention, leading to calls for urgent community discussion of the ethical issues involved. LLMs generate text by representing and predicting statistical properties of language. Optimized for statistical patterns and linguistic form rather than for

truth or reliability, these models cannot assess the quality of the information they use.

Recent work has highlighted ethical risks that are associated with LLMs, including biases that arise from training data; environmental and socioeconomic impacts; privacy and confidentiality risks; the perpetuation of stereotypes; and the potential for deliberate or accidental misuse. We focus on a distinct set of ethical questions concerning moral responsibility—specifically blame and credit—for LLM-generated

content. We argue that different responsibility standards apply to positive and negative uses (or outputs) of LLMs and offer preliminary recommendations. These include: calls for updated guidance from policymakers that reflect this asymmetry in responsibility standards; transparency norms; technology goals; and the establishment of interactive forums for participatory debate on LLMs.‌

(cut)

Credit–blame asymmetry may lead to achievement gaps

Since the Industrial Revolution, automating technologies have made workers redundant in many industries, particularly in agriculture and manufacturing. The recent assumption25 has been that creatives

and knowledge workers would remain much less impacted by these changes in the near-to-mid-term future. Advances in LLMs challenge this premise.

How these trends will impact human workforces is a key but unresolved question. The spread of AI-based applications and tools such as LLMs will not necessarily replace human workers; it may simply

shift them to tasks that complement the functions of the AI. This may decrease opportunities for human beings to distinguish themselves or excel in workplace settings. Their future tasks may involve supervising or maintaining LLMs that produce the sorts of outputs (for example, text or recommendations) that skilled human beings were previously producing and for which they were receiving credit. Consequently, work in a world relying on LLMs might often involve ‘achievement gaps’ for human beings: good, useful outcomes will be produced, but many of them will not be achievements for which human workers and professionals can claim credit.

This may result in an odd state of affairs. If responsibility for positive and negative outcomes produced by LLMs is asymmetrical as we have suggested, humans may be justifiably held responsible for negative outcomes created, or allowed to happen, when they or their organizations make use of LLMs. At the same time, they may deserve less credit for AI-generated positive outcomes, as they may not be displaying the skills and talents needed to produce text, exerting judgment to make a recommendation, or generating other creative outputs.

Can AI language models replace human participants?

Dillon, D, Tandon, N., Gu, Y., & Gray, K.

Trends in Cognitive Sciences

May 10, 2023

Abstract

Recent work suggests that language models such as GPT can make human-like judgments across a number of domains. We explore whether and when language models might replace human participants in psychological science. We review nascent research, provide a theoretical model, and outline caveats of using AI as a participant.

(cut)

Does GPT make human-like judgments?

We initially doubted the ability of LLMs to capture human judgments but, as we detail in Box 1, the moral judgments of GPT-3.5 were extremely well aligned with human moral judgments in our analysis (r= 0.95;

full details at https://nikett.github.io/gpt-as-participant). Human morality is often argued to be especially difficult for language models to capture and yet we found powerful alignment between GPT-3.5 and human judgments.

We emphasize that this finding is just one anecdote and we do not make any strong claims about the extent to which LLMs make human-like judgments, moral or otherwise. Language models also might be especially good at predicting moral judgments because moral judgments heavily hinge on the structural features of scenarios, including the presence of an intentional agent, the causation of damage, and a vulnerable victim, features that language models may have an easy time detecting.  However, the results are intriguing.

Other researchers have empirically demonstrated GPT-3’s ability to simulate human participants in domains beyond moral judgments, including predicting voting choices, replicating behavior in economic games, and displaying human-like problem solving and heuristic judgments on scenarios from cognitive

psychology. LLM studies have also replicated classic social science findings including the Ultimatum Game and the Milgram experiment. One company (http://syntheticusers.com) is expanding on these

findings, building infrastructure to replace human participants and offering ‘synthetic AI participants’

for studies.

(cut)

From Caveats and looking ahead

Language models may be far from human, but they are trained on a tremendous corpus of human expression and thus they could help us learn about human judgments. We encourage scientists to compare simulated language model data with human data to see how aligned they are across different domains and populations.  Just as language models like GPT may help to give insight into human judgments, comparing LLMs with human judgments can teach us about the machine minds of LLMs; for example, shedding light on their ethical decision making.

Lurking under the specific concerns about the usefulness of AI language models as participants is an age-old question: can AI ever be human enough to replace humans? On the one hand, critics might argue that AI participants lack the rationality of humans, making judgments that are odd, unreliable, or biased. On the other hand, humans are odd, unreliable, and biased – and other critics might argue that AI is just too sensible, reliable, and impartial.  What is the right mix of rational and irrational to best capture a human participant?  Perhaps we should ask a big sample of human participants to answer that question. We could also ask GPT.

The research is here.

Are We Ready for AI to Raise the Dead?

Jack Holmes

Esquire Magazine

Originally posted 4 May 24

Here is an excerpt:

You can see wonderful possibilities here. Some might find comfort in hearing their mom’s voice, particularly if she sounds like she really sounded and gives the kind of advice she really gave. But Sandel told me that when he presents the choice to students in his ethics classes, the reaction is split, even as he asks in two different ways. First, he asks whether they’d be interested in the chatbot if their loved one bequeathed it to them upon their death. Then he asks if they’d be interested in building a model of themselves to bequeath to others. Oh, and what if a chatbot is built without input from the person getting resurrected? The notion that someone chose to be represented posthumously in a digital avatar seems important, but even then, what if the model makes mistakes? What if it misrepresents—slanders, even—the dead?

Soon enough, these questions won’t be theoretical, and there is no broad agreement about whom—or even what—to ask. We’re approaching a more fundamental ethical quandary than we often hear about in discussions around AI: human bias embedded in algorithms, privacy and surveillance concerns, mis- and disinformation, cheating and plagiarism, the displacement of jobs, deepfakes. These issues are really all interconnected—Osama bot Laden might make the real guy seem kinda reasonable or just preach jihad to tweens—and they all need to be confronted. We think a lot about the mundane (kids cheating in AP History) and the extreme (some advanced AI extinguishing the human race), but we’re more likely to careen through the messy corridor in between. We need to think about what’s allowed and how we’ll decide.

(cut)

Our governing troubles are compounded by the fact that, while a few firms are leading the way on building these unprecedented machines, the technology will soon become diffuse. More of the codebase for these models is likely to become publicly available, enabling highly talented computer scientists to build their own in the garage. (Some folks at Stanford have already built a ChatGPT imitator for around $600.) What happens when some entrepreneurial types construct a model of a dead person without the family’s permission? (We got something of a preview in April when a German tabloid ran an AI-generated interview with ex–Formula 1 driver Michael Schumacher, who suffered a traumatic brain injury in 2013. His family threatened to sue.) What if it’s an inaccurate portrayal or it suffers from what computer scientists call “hallucinations,” when chatbots spit out wildly false things? We’ve already got revenge porn. What if an old enemy constructs a false version of your dead wife out of spite? “There’s an important tension between open access and safety concerns,” Reich says. “Nuclear fusion has enormous upside potential,” too, he adds, but in some cases, open access to the flesh and bones of AI models could be like “inviting people around the world to play with plutonium.”

The article is here.

Yes, there was a Black Mirror episode (Be Right Back) about this issue.  The wiki is here.

Fighting for our cognitive liberty

Liz Mineo

The Harvard Gazette

Originally published 26 April 23

Imagine going to work and having your employer monitor your brainwaves to see whether you’re mentally tired or fully engaged in filling out that spreadsheet on April sales.

Nita Farahany, professor of law and philosophy at Duke Law School and author of “The Battle for Your Brain: Defending the Right to Think Freely in the Age of Neurotechnology,” says it’s already happening, and we all should be worried about it.

Farahany highlighted the promise and risks of neurotechnology in a conversation with Francis X. Shen, an associate professor in the Harvard Medical School Center for Bioethics and the MGH Department of Psychiatry, and an affiliated professor at Harvard Law School. The Monday webinar was co-sponsored by the Harvard Medical School Center for Bioethics, the Petrie-Flom Center for Health Law Policy, Biotechnology, and Bioethics at Harvard Law School, and the Dana Foundation.

Farahany said the practice of tracking workers’ brains, once exclusively the stuff of science fiction, follows the natural evolution of personal technology, which has normalized the use of wearable devices that chronicle heartbeats, footsteps, and body temperatures. Sensors capable of detecting and decoding brain activity already have been embedded into everyday devices such as earbuds, headphones, watches, and wearable tattoos.

“Commodification of brain data has already begun,” she said. “Brain sensors are already being sold worldwide. It isn’t everybody who’s using them yet. When it becomes an everyday part of our everyday lives, that’s the moment at which you hope that the safeguards are already in place. That’s why I think now is the right moment to do so.”

Safeguards to protect people’s freedom of thought, privacy, and self-determination should be implemented now, said Farahany. Five thousand companies around the world are using SmartCap technologies to track workers’ fatigue levels, and many other companies are using other technologies to track focus, engagement and boredom in the workplace.

If protections are put in place, said Farahany, the story with neurotechnology could be different than the one Shoshana Zuboff warns of in her 2019 book, “The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power.” In it Zuboff, Charles Edward Wilson Professor Emerita at Harvard Business School, examines the threat of the widescale corporate commodification of personal data in which predictions of our consumer activities are bought, sold, and used to modify behavior.

The article is here.

Artificial intelligence, superefficiency and the end of work: a humanistic perspective on meaning in life

Knell, S., Rüther, M.

AI Ethics (2023).

https://doi.org/10.1007/s43681-023-00273-w

Abstract

How would it be assessed from an ethical point of view if human wage work were replaced by artificially intelligent systems (AI) in the course of an automation process? An answer to this question has been discussed above all under the aspects of individual well-being and social justice. Although these perspectives are important, in this article, we approach the question from a different perspective: that of leading a meaningful life, as understood in analytical ethics on the basis of the so-called meaning-in-life debate. Our thesis here is that a life without wage work loses specific sources of meaning, but can still be sufficiently meaningful in certain other ways. Our starting point is John Danaher’s claim that ubiquitous automation inevitably leads to an achievement gap. Although we share this diagnosis, we reject his provocative solution according to which game-like virtual realities could be an adequate substitute source of meaning. Subsequently, we outline our own systematic alternative which we regard as a decidedly humanistic perspective. It focuses both on different kinds of social work and on rather passive forms of being related to meaningful contents. Finally, we go into the limits and unresolved points of our argumentation as part of an outlook, but we also try to defend its fundamental persuasiveness against a potential objection.

From Concluding remarks

In this article, we explored the question of how we can find meaning in a post-work world. Our answer relies on a critique of John Danaher’s utopia of games and tries to stick to the humanistic idea, namely to the idea that we do not have to alter our human lifeform in an extensive way and also can keep up our orientation towards common ideals, such as working towards the good, the true and the beautiful.

Our proposal still has some shortcomings, which include the following two that we cannot deal with extensively but at least want to briefly comment on. First, we assumed that certain professional fields, especially in the meaning conferring area of the good, cannot be automated, so that the possibility of mini-jobs in these areas can be considered. This assumption is based on a substantial thesis from the philosophy of mind, namely that AI systems cannot develop consciousness and consequently also no genuine empathy. This assumption needs to be further elaborated, especially in view of some forecasts that even the altruistic and philanthropic professions are not immune to the automation of superefficient systems. Second, we have adopted without further critical discussion the premise of the hybrid standard model of a meaningful life according to which meaning conferring objective value is to be found in the three spheres of the true, the good, and the beautiful. We take this premise to be intuitively appealing, but a further elaboration of our argumentation would have to try to figure out, whether this trias is really exhaustive, and if so, due to which underlying more general principle. Third, the receptive side of finding meaning in the realm of the true and beautiful was emphasized and opposed to the active striving towards meaningful aims. Here, we have to more precisely clarify what axiological status reception has in contrast to active production—whether it is possibly meaning conferring to a comparable extent or whether it is actually just a less meaningful form. This is particularly important to be able to better assess the appeal of our proposal, which depends heavily on the attractiveness of the vita contemplativa.

The article is here.

Doctors are drowning in paperwork. Some companies claim AI can help

Geoff Brumfiel

NPR.org - Health Shots

Originally posted 5 APR 23

Here are two excerpts:

But Paul kept getting pinged from younger doctors and medical students. They were using ChatGPT, and saying it was pretty good at answering clinical questions. Then the users of his software started asking about it.

In general, doctors should not be using ChatGPT by itself to practice medicine, warns Marc Succi, a doctor at Massachusetts General Hospital who has conducted evaluations of how the chatbot performs at diagnosing patients. When presented with hypothetical cases, he says, ChatGPT could produce a correct diagnosis accurately at close to the level of a third- or fourth-year medical student. Still, he adds, the program can also hallucinate findings and fabricate sources.

"I would express considerable caution using this in a clinical scenario for any reason, at the current stage," he says.

But Paul believed the underlying technology can be turned into a powerful engine for medicine. Paul and his colleagues have created a program called "Glass AI" based off of ChatGPT. A doctor tells the Glass AI chatbot about a patient, and it can suggest a list of possible diagnoses and a treatment plan. Rather than working from the raw ChatGPT information base, the Glass AI system uses a virtual medical textbook written by humans as its main source of facts – something Paul says makes the system safer and more reliable.

(cut)

Nabla, which he co-founded, is now testing a system that can, in real time, listen to a conversation between a doctor and a patient and provide a summary of what the two said to one another. Doctors inform their patients that the system is being used in advance, and as a privacy measure, it doesn't actually record the conversation.

"It shows a report, and then the doctor will validate with one click, and 99% of the time it's right and it works," he says.

The summary can be uploaded to a hospital records system, saving the doctor valuable time.

Other companies are pursuing a similar approach. In late March, Nuance Communications, a subsidiary of Microsoft, announced that it would be rolling out its own AI service designed to streamline note-taking using the latest version of ChatGPT, GPT-4. The company says it will showcase its software later this month.

The info is here.

Foundation Models are exciting, but they should not disrupt the foundations of caring

Morley, Jessica and Floridi, Luciano

(April 20, 2023).

Available at SSRN: https://ssrn.com/abstract=

Abstract

The arrival of Foundation Models in general, and Large Language Models (LLMs) in particular, capable of ‘passing’ medical qualification exams at or above a human level, has sparked a new wave of ‘the chatbot will see you now’ hype. It is exciting to witness such impressive technological progress, and LLMs have the potential to benefit healthcare systems, providers, and patients. However, these benefits are unlikely to be realised by propagating the myth that, just because LLMs are sometimes capable of passing medical exams, they will ever be capable of supplanting any of the main diagnostic, prognostic, or treatment tasks of a human clinician. Contrary to popular discourse, LLMs are not necessarily more efficient, objective, or accurate than human healthcare providers. They are vulnerable to errors in underlying ‘training’ data and prone to ‘hallucinating’ false information rather than facts. Moreover, there are nuanced, qualitative, or less measurable reasons why it is prudent to be mindful of hyperbolic claims regarding the transformative power ofLLMs. Here we discuss these reasons, including contextualisation, empowerment, learned intermediaries, manipulation, and empathy. We conclude that overstating the current potential of LLMs does a disservice to the complexity of healthcare and the skills of healthcare practitioners and risks a ‘costly’ new AI winter. A balanced discussion recognising the potential benefits and limitations can help avoid this outcome.

Conclusion

The technical feats achieved by foundation models in the last five years, and especially in the last six months, are undeniably impressive. Also undeniable is the fact that most healthcare systems across the world are under considerable strain. It is right, therefore, to recognise and invest in the potentially transformative power of models such as Med-PaLM and ChatGPT – healthcare systems will almost certainly benefit.  However, overstating their current potential does a disservice to the complexity of healthcare and the skills required of healthcare practitioners. Not only does this ‘hype’ risk direct patient and societal harm, but it also risks re-creating the conditions of previous AI winters when investors and enthusiasts became discouraged by technological developments that over-promised and under-delivered. This could be the most harmful outcome of all, resulting in significant opportunity costs and missed chances to benefit transform healthcare and benefit patients in smaller, but more positively impactful, ways. A balanced approach recognising the potential benefits and limitations can help avoid this outcome. 

A Psychologist Explains How AI and Algorithms Are Changing Our Lives

Danny Lewis

The Wall Street Journal

Originally posted 21 MAR 23

In an age of ChatGPT, computer algorithms and artificial intelligence are increasingly embedded in our lives, choosing the content we’re shown online, suggesting the music we hear and answering our questions.

These algorithms may be changing our world and behavior in ways we don’t fully understand, says psychologist and behavioral scientist Gerd Gigerenzer, the director of the Harding Center for Risk Literacy at the University of Potsdam in Germany. Previously director of the Center for Adaptive Behavior and Cognition at the Max Planck Institute for Human Development, he has conducted research over decades that has helped shape understanding of how people make choices when faced with uncertainty. 

In his latest book, “How to Stay Smart in a Smart World,” Dr. Gigerenzer looks at how algorithms are shaping our future—and why it is important to remember they aren’t human. He spoke with the Journal for The Future of Everything podcast.

The term algorithm is thrown around so much these days. What are we talking about when we talk about algorithms?

It is a huge thing, and therefore it is important to distinguish what we are talking about. One of the insights in my research at the Max Planck Institute is that if you have a situation that is stable and well defined, then complex algorithms such as deep neural networks are certainly better than human performance. Examples are [the games] chess and Go, which are stable. But if you have a problem that is not stable—for instance, you want to predict a virus, like a coronavirus—then keep your hands off complex algorithms. [Dealing with] the uncertainty—that is more how the human mind works, to identify the one or two important cues and ignore the rest. In that type of ill-defined problem, complex algorithms don’t work well. I call this the “stable world principle,” and it helps you as a first clue about what AI can do. It also tells you that, in order to get the most out of AI, we have to make the world more predictable.

So after all these decades of computer science, are algorithms really just still calculators at the end of the day, running more and more complex equations?

What else would they be? A deep neural network has many, many layers, but they are still calculating machines. They can do much more than ever before with the help of video technology. They can paint, they can construct text. But that doesn’t mean that they understand text in the sense humans do.

The article is here.

We need an AI rights movement

Jacy Reese Anthis

The Hill

Originally posted 23 MAR 23

New artificial intelligence technologies like the recent release of GPT-4 have stunned even the most optimistic researchers. Language transformer models like this and Bing AI are capable of conversations that feel like talking to a human, and image diffusion models such as Midjourney and Stable Diffusion produce what looks like better digital art than the vast majority of us can produce. 

It’s only natural, after having grown up with AI in science fiction, to wonder what’s really going on inside the chatbot’s head. Supporters and critics alike have ruthlessly probed their capabilities with countless examples of genius and idiocy. Yet seemingly every public intellectual has a confident opinion on what the models can and can’t do, such as claims from Gary Marcus, Judea Pearl, Noam Chomsky, and others that the models lack causal understanding.

But thanks to tools like ChatGPT, which implements GPT-4, being publicly accessible, we can put these claims to the test. If you ask ChatGPT why an apple falls, it gives a reasonable explanation of gravity. You can even ask ChatGPT what happens to an apple released from the hand if there is no gravity, and it correctly tells you the apple will stay in place. 

Despite these advances, there seems to be consensus at least that these models are not sentient. They have no inner life, no happiness or suffering, at least no more than an insect. 

But it may not be long before they do, and our concepts of language, understanding, agency, and sentience are deeply insufficient to assess the AI systems that are becoming digital minds integrated into society with the capacity to be our friends, coworkers, and — perhaps one day — to be sentient beings with rights and personhood. 

AIs are no longer mere tools like smartphones and electric cars, and we cannot treat them in the same way as mindless technologies. A new dawn is breaking. 

This is just one of many reasons why we need to build a new field of digital minds research and an AI rights movement to ensure that, if the minds we create are sentient, they have their rights protected. Scientists have long proposed the Turing test, in which human judges try to distinguish an AI from a human by speaking to it. But digital minds may be too strange for this approach to tell us what we need to know. 

The info is here.

Why artificial intelligence needs to understand consequences

Neil Savage

Nature

Originally published 24 FEB 23

Here is an excerpt:

The headline successes of AI over the past decade — such as winning against people at various competitive games, identifying the content of images and, in the past few years, generating text and pictures in response to written prompts — have been powered by deep learning. By studying reams of data, such systems learn how one thing correlates with another. These learnt associations can then be put to use. But this is just the first rung on the ladder towards a loftier goal: something that Judea Pearl, a computer scientist and director of the Cognitive Systems Laboratory at the University of California, Los Angeles, refers to as “deep understanding”.

In 2011, Pearl won the A.M. Turing Award, often referred to as the Nobel prize for computer science, for his work developing a calculus to allow probabilistic and causal reasoning. He describes a three-level hierarchy of reasoning4. The base level is ‘seeing’, or the ability to make associations between things. Today’s AI systems are extremely good at this. Pearl refers to the next level as ‘doing’ — making a change to something and noting what happens. This is where causality comes into play.

A computer can develop a causal model by examining interventions: how changes in one variable affect another. Instead of creating one statistical model of the relationship between variables, as in current AI, the computer makes many. In each one, the relationship between the variables stays the same, but the values of one or several of the variables are altered. That alteration might lead to a new outcome. All of this can be evaluated using the mathematics of probability and statistics. “The way I think about it is, causal inference is just about mathematizing how humans make decisions,” Bhattacharya says.

Bengio, who won the A.M. Turing Award in 2018 for his work on deep learning, and his students have trained a neural network to generate causal graphs5 — a way of depicting causal relationships. At their simplest, if one variable causes another variable, it can be shown with an arrow running from one to the other. If the direction of causality is reversed, so too is the arrow. And if the two are unrelated, there will be no arrow linking them. Bengio’s neural network is designed to randomly generate one of these graphs, and then check how compatible it is with a given set of data. Graphs that fit the data better are more likely to be accurate, so the neural network learns to generate more graphs similar to those, searching for one that fits the data best.

This approach is akin to how people work something out: people generate possible causal relationships, and assume that the ones that best fit an observation are closest to the truth. Watching a glass shatter when it is dropped it onto concrete, for instance, might lead a person to think that the impact on a hard surface causes the glass to break. Dropping other objects onto concrete, or knocking a glass onto a soft carpet, from a variety of heights, enables a person to refine their model of the relationship and better predict the outcome of future fumbles.

The info is here.

White Supremacist Networks Gab and 8Kun Are Training Their Own AI Now

David Gilbert

Vice News

Originally posted 22 FEB 23

Here are two excerpts:

Artificial intelligence is everywhere right now, and many are questioning the safety and morality of the AI systems released by some of the world’s biggest companies, including Open AI’s ChatGPT, Bing’s Sydney, and Google’s Bard. It was only a matter of time until the online spaces where extremists gather became interested in the technology.

Gab is a social network filled with homophobic, christian nationalist and white supremacist content. On Tuesday its CEO Andrew Torba announced the launch of its AI image generator, Gabby.

“At Gab, we have been experimenting with different AI systems that have popped up over the past year,” Torba wrote in a statement. “Every single one is skewed with a liberal/globalist/talmudic/satanic worldview. What if Gab AI Inc builds a Gab .ai (see what I did there?) that is based, has no ‘hate speech” filters and doesn’t obfuscate and distort historical and Biblical Truth?”

Gabby is currently live on Gab’s site and available to all members. Like Midjourney and DALL-E, it is an image generator that users interact with by sending it a prompt, and within seconds it will generate entirely new images based on that prompt.

Echoing his past criticisms of Big Tech platforms like Facebook and Twitter, Torba claims that mainstream platforms are now “censoring” their AI systems to prevent people from discussing right-wing topics such as Christian nationalism. Torba’s AI, by contrast, will have ”the ability to speak freely without the constraints of liberal propaganda wrapped tightly around its neck.”

(cut)

8chan, which was founded to support the Gamergate movement, became the home of QAnon in early 2018 and was taken offline in August 2019 after the man who killed 20 people at an El Paso Walmart posted an anti-immigrant screed on the site.

Watkins has been speaking about his AI system for a few weeks now, but has yet to reveal how it will work or when it will launch. Watkins’ central selling point, like Torba’s, appears to be that his system will be “uncensored.”

“So that we can compete against these people that are putting up all of these false flags and illusions,” Watkins said on Feb. 13 when he was asked why he was creating an AI system.  “We are working on our own AI that is going to give you an uncensored look at the way things are going,” Watkins said in a video interview at the end of January.But based on some of the images the engine is churning out, Watkins still has a long way to go to perfect his AI image generator.

The info is here.

What Happens When AI Has Read Everything?

Ross Anderson

The Atlantic

Originally posted 18 JAN 23

Here is an excerpt:

Ten trillion words is enough to encompass all of humanity’s digitized books, all of our digitized scientific papers, and much of the blogosphere. That’s not to say that GPT-4 will have read all of that material, only that doing so is well within its technical reach. You could imagine its AI successors absorbing our entire deep-time textual record across their first few months, and then topping up with a two-hour reading vacation each January, during which they could mainline every book and scientific paper published the previous year.

Just because AIs will soon be able to read all of our books doesn’t mean they can catch up on all of the text we produce. The internet’s storage capacity is of an entirely different order, and it’s a much more democratic cultural-preservation technology than book publishing. Every year, billions of people write sentences that are stockpiled in its databases, many owned by social-media platforms.

Random text scraped from the internet generally doesn’t make for good training data, with Wikipedia articles being a notable exception. But perhaps future algorithms will allow AIs to wring sense from our aggregated tweets, Instagram captions, and Facebook statuses. Even so, these low-quality sources won’t be inexhaustible. According to Villalobos, within a few decades, speed-reading AIs will be powerful enough to ingest hundreds of trillions of words—including all those that human beings have so far stuffed into the web.

And the conclusion:

If, however, our data-gorging AIs do someday surpass human cognition, we will have to console ourselves with the fact that they are made in our image. AIs are not aliens. They are not the exotic other. They are of us, and they are from here. They have gazed upon the Earth’s landscapes. They have seen the sun setting on its oceans billions of times. They know our oldest stories. They use our names for the stars. Among the first words they learn are flow, mother, fire, and ash.

The article is here.

Growth of AI in mental health raises fears of its ability to run wild

Sabrina Moreno

Axios.com

Originally posted 9 MAR 23

Here's how it begins:

The rise of AI in mental health care has providers and researchers increasingly concerned over whether glitchy algorithms, privacy gaps and other perils could outweigh the technology's promise and lead to dangerous patient outcomes.

Why it matters: As the Pew Research Center recently found, there's widespread skepticism over whether using AI to diagnose and treat conditions will complicate a worsening mental health crisis.

• Mental health apps are also proliferating so quickly that regulators are hard-pressed to keep up.
• The American Psychiatric Association estimates there are more than 10,000 mental health apps circulating on app stores. Nearly all are unapproved.

What's happening: AI-enabled chatbots like Wysa and FDA-approved apps are helping ease a shortage of mental health and substance use counselors.

• The technology is being deployed to analyze patient conversations and sift through text messages to make recommendations based on what we tell doctors.
• It's also predicting opioid addiction risk, detecting mental health disorders like depression and could soon design drugs to treat opioid use disorder.

Driving the news: The fear is now concentrated around whether the technology is beginning to cross a line and make clinical decisions, and what the Food and Drug Administration is doing to prevent safety risks to patients.

• KoKo, a mental health nonprofit, recently used ChatGPT as a mental health counselor for about 4,000 people who weren't aware the answers were generated by AI, sparking criticism from ethicists.
• Other people are turning to ChatGPT as a personal therapist despite warnings from the platform saying it's not intended to be used for treatment.

The info is here.