Is AI Mystical? A False God? The Latest Tech-Bro Hype?

The stock markets recently experienced a dip, with concerns over the prospects for the current artificial intelligence (AI) boom playing a role. As of the end of July:

Investors are sending mixed signals regarding their appetite for tech stocks, as the growing debate over the artificial intelligence boom, and a US clampdown on chip exports to China, raise questions over the direction of growth for key companies.

There were fears of a fresh sell-off after the US-listed shares in the chip maker Nvidia dropped 7% overnight, amid concerns that excitement over companies at the forefront of AI development had been overblown.

Nvidia, which has been the biggest beneficiary of the AI boom, has now fallen by 20% from last month’s peak. The semiconductor designer Arm, which has also benefited from the AI hype, ended the session down 6%.

Shares in Microsoft were down almost 3% in after-hours trading, after it revealed that growth at its cloud division, Azure, had slowed as it struggled to keep up with AI-related demands. (1)

“Tech bubbles” have been happening for a long time. Longer than the phrase “Silicon Valley” has been around. When railroads were still at the cutting edge of innovative technology, When the market for speculative investments by the Jay Cooke & Company bank in Northern Pacific Railway bonds collapsed, it set off what is now remember as the Panic of 1873. Which quickly spread to Europe and other parts of the world. (2)

Navneet Alang in The Walrus in May published some reflections on the AI hype, with a focus on the large language models (LLM) used by chatbots. (3) He touches on a number of issues, including how to define AI “intelligence” in relation to the human kind and how definitions of what intelligence, thinking, mind, self-awareness, and meaning are more than a little complicated just for the human kinds. (The Walrus headline is “AI Is a False God.”)

Sometimes the line between explaining an impression of what AI is about and promoting confusing concepts about what it is can be very blurred. Alang writes:

The sense of there being a thinking thing behind AI chatbots is also driven by the now common wisdom that we don’t know exactly how AI systems work. What’s called the black box problem is often framed as mysticism—the robots are so far ahead or so alien that they are doing something we can’t comprehend. That is true, but not quite in the way it sounds. New York University professor Leif Weatherby suggests that the models are processing so many permutations of data that it is impossible for a single person to wrap their head around it. The mysticism of AI isn’t a hidden or inscrutable mind behind the curtain; it’s to do with scale and brute power. [my emphasis]

Mysticism here is more-or-less equated with something that’s not easy to understand. There’s nothing that exotic about the notion that computers can perform lots of calculations on data and do it faster than humans working with spreadsheets and pencils is not exactly new. That kind of, you know, what computers do.

Alang’s description of attending a Microsoft event earlier this year meant to tout the wonders of AI – i.e., to promote the investor buzz around the whole field - is interesting and entertaining. But, despite the fact that he writes as though he is surprised that tech boosters promise utopian results from their latest program or gadget – not exactly a new development! he provides an important qualification to the undifferentiated hype:

Yet for all the high-minded talk of what AI might one day do, much of what artificial intelligence appeared to do best was entirely quotidian: taking financial statements and reconciling figures, making security practices more responsive and efficient, transcribing and summarizing meetings, triaging emails more efficiently. What that emphasis on day-to-day tasks suggested is that AI isn’t so much going to produce a grand new world as, depending on your perspective, make what exists now slightly more efficient — or, rather, intensify and solidify the structure of the present. Yes, some parts of your job might be easier, but what seems likely is that those automated tasks will in turn simply be part of more work. [my emphasis]

That’s a helpful perspective, as long we recognize that a lot of “slightly more efficient” starts to look like a “new world” after a while, though not necessarily a grander one.

He also tells a good when-accomplished-historian-meets-a-famous-tech-bro-oligarch story:

Some Silicon Valley businessmen have taken tech solutionism to an extreme.

Actually, taking “tech solutionism to an extreme” is almost a synonym for Silicon Valley. An academic friend of mine in the Bay Area told me about being somewhat bewildered at some of her techie acquaintances who would talk enthusiastically about the “revolutionary” new approach they were coming up with. But then when you got to specifics about the program they working on – it was a new video game. With bizillions of dollars in venture capital chasing new possibilities, that kind of hype becomes part of daily life. The venture capitalists, though, are working with a model in which a large portion of the startups in which they invest will never turn out to be profitable. Their business model is to make money on the ones that are. So tech entrepreneurs learn how to make convincing pitches. Continuing with the story:

It is these AI accelerationists whose ideas are the most terrifying. Marc Andreessen was intimately involved in the creation of the first web browsers and is now a billionaire venture capitalist who has taken up a mission to fight against the “woke mind virus” and generally embrace capitalism and libertarianism. In a screed published last year, titled “The Techno-Optimist Manifesto,” Andreessen outlined his belief that “there is no material problem—whether created by nature or by technology—that cannot be solved with more technology.” When writer Rick Perlstein attended a dinner at Andreessen’s $34 million (US) home in California, he found a group adamantly opposed to regulation or any kind of constraint on tech (in a tweet at the end of 2023, Andreessen called regulation of AI “the new foundation of totalitarianism”). When Perlstein related the whole experience to a colleague, he “noted a similarity to a student of his who insisted that all the age-old problems historians worried over would soon obviously be solved by better computers, and thus considered the entire humanistic enterprise faintly ridiculous.”

There is no shortage of people who confuse “technology” with “magic.” Like the magical solutions to climate change that will somehow magically be produced by Technology – an argument advanced by groups like oil lobbyists who don’t want any solutions to problems that don’t involve relying on fossil fuels from now to eternity. Actual, practical, non-magical solutions like batteries, wind farms, and solar panels don’t count as the utopian answer that the mysterious force known as Technology always has on the horizon – just not right now.

It's worth noting also that some tech oligarchs also use the AI Doomsday talk to promote regulations that they prefer, not because the Skynet of the Terminator movies is about to take over, but because they want to restrict competition that might dilute their monopoly market control. It’s always a good idea to pay attention to the oligarch behind the curtain. (4)

Alang offers a decent framework for parsing superficial “tech” fantasies:

A common understanding of technology is that it is a tool. You have a task you need to do, and tech helps you accomplish it. But there are some significant tech­nologies—shelter, the printing press, the nuclear bomb or the ­rocket, the internet—that almost “re-render” the world and thus change something about how we conceive of both ourselves and reality. It’s not a mere evolution. After the arrival of the book, and with it the capacity to document complex knowledge and disseminate information outside of the established gatekeepers, the ground of reality itself changed.

AI occupies a strange position, in that it likely represents one of those sea changes in technology but is at the same time overhyped. The idea that AI will lead us to some grand utopia is deeply flawed. Technology does, in fact, turn over new ground, but what was there in soil doesn’t merely go away.

Large language models are also a long way from Mr. Data levels of thinking and learning proficiency. A recent Nature article discusses the risk of “model collapse” for LLMs, i.e., “what may happen to GPT-{n} [chatbots] once LLMs contribute much of the text found online. We find that indiscriminate use of model-generated content in training causes irreversible defects in the resulting models, in which tails of the original content distribution disappear.” (5)

Alang’s article is a helpful Big Picture look at the current AI discussion.

But while no article or book can cover every nuance of a topic like this, I would stress a couple of other features of AI that are important to keep in mind. One is that AI computations and syntheses of information are not modelled on the human brain, although some developments in AI have given scientists some better insight on how human brains function.

The other is the energy problem that billionaire tech bros would mostly like to dismiss, because it disturbs their fantasies about achieving immortality by transferring their brains’ software into androids or whatever. AI consumes a tremendous amount of energy compared to human brains. The biological versions are orders of magnitude more energy-efficient than the current AI devices.

Notes:

(1) Makortoff, Kalyeena & Jolly, Jasper (2024): Mixed signals on tech stocks amid debate over viability of AI boom. The Guardian 07/31/2024. <https://www.theguardian.com/business/article/2024/jul/31/mixed-signals-on-tech-stocks-amid-debate-over-viability-of-ai-boom> (Accessed: 2024-08-08).

(2) Butts, Mickey (2023): How One Robber Baron’s Gamble on Railroads Brought Down His Bank and Plunged the U.S. Into the First Great Depression. Smithsonian Magazine 09/18/2024. <https://www.smithsonianmag.com/history/robber-baron-gamble-railroads-brought-down-bank-plunged-us-into-first-great-depression-panic-1873-180982877/> (Accessed: 2024-08-08).

(3) Alang, Navneet (2024): AI Is a False God. The Walrus 05/29/2024. <https://thewalrus.ca/ai-hype/> (Accessed: 2024-08-08).

(4) Pay no attention to that man behind the curtain. GreyAllen7 YouTube channel 11/30/2006. <https://youtu.be/YWyCCJ6B2WE?si=-iqT-zaFpCpNjoXV> (Accessed: 2024-08-08).

(5) Shumailov, Ilia et. al. (2024): AI models collapse when trained on recursively generated data. Nature 07/24/2024. <https://www.nature.com/articles/s41586-024-07566-y> (Accessed: 2024-08-08).

AI risks: real, exaggerated, and imaginary

One of the striking features of the current AI “wave” is that it's often the same people spinning the techno-utopian dreams who also hype the AI doomerism. Tech-bro billionaires like Sam Alman or Peter Thiel who have their various techno-utopian fantasies are also calling for more regulation to head off the supposedly dystopian consequences that could result from AI developments.

Alex Hann and Emily Bender have a helpful take on this. Which is that it’s a strategy by the corporate AI boosters and their hangers-on to divert public and regulatory attention from the real existing negative consequences that are actually visible today. (1)

Artificial intelligence – the term itself was established in 1956 (2) - definitely has positive uses in medicine and all kinds of research applications. But - for better or worse - we're still a long way from Mister Data androids or Skynet.

Biological research on animal communication, for instance, is an intriguing application that is already being used:

Whether animals communicate with one another in terms we might be able to understand is a question of enduring fascination. Although people in many Indigenous cultures have long believed that animals can intentionally communicate, Western scientists traditionally have shied away from research that blurs the lines between humans and other animals for fear of being accused of anthropomorphism. But with recent breakthroughs in AI, “people realize that we are on the brink of fairly major advances in regard to understanding animals' communicative behavior,” Rutz says.

Beyond creating chatbots that woo people and producing art that wins fine-arts competitions, machine learning may soon make it possible to decipher things like crow calls, says Aza Raskin, one of the founders of the nonprofit Earth Species Project. Its team of artificial-intelligence scientists, biologists and conservation experts is collecting a wide range of data from a variety of species and building machine-learning models to analyze them. Other groups such as the Project Cetacean Translation Initiative (CETI) are focusing on trying to understand a particular species, in this case the sperm whale.

Decoding animal vocalizations could aid conservation and welfare efforts. (3)

Aside from the real and much-discussed possibilities for Deep Fakes and cheating on school and college work, there are two other big dark sides that we see in the real world for the current AI wave. One is that even since before the term AI was invented, the technologies that are part of it have been largely driven by defense needs, much of the basic research done with public funds especially by DARPA, the Pentagon tech-lab agency. There will also be a market for new whiz-bang technology that can be marketed as weapons that kill more efficiently.

AI and the military

Defense One reports:

The U.S. Army plans to ask contractors for help integrating industry-generated artificial-intelligence algorithms into its operations, part of the service's 100-day push to lay the groundwork for sweeping adoption of AI.

“One of the things that we want to do is we want to adopt third-party-generated algorithms as fast as y'all are building,” Young Bang, the principal deputy assistant Army secretary for acquisition, logistics and technology, said at the Amazon Web Services Washington D.C. Summit on Wednesday. “We realized, while we had tons of data…we're not gonna develop our algorithms better than y’all.”

Bang said that as one of the largest consumers of AI and algorithm technologies, the Army is anxious to generate partnerships with industry and incorporate newer proprietary technology.

“We want to have a partnership. We're going to break down obstacles for us to adopt third-party-generated algorithms,” he said. [my emphasis] (4)

The Israeli Defense Forces provided the world a couple of dramatic and grim examples with the “Lavender” and “Where’s Daddy?” AI military technology. (5) Neither of these are autonomous of their human users, of course. Military AI isn’t run by Skynet or some android. Human being build them, employ them, and bear responsibility for their misuse.

The dark side of science and technology is nothing new. They can be used for good or bad and has always been used for both, just as a hammer can be used to drive nails or to hit someone in the head. Oe the one hand, that’s a fairly banal observation. But it’s also important to keep constantly in mind, maybe especially when looking at tech fads like cryptocurrency or AI.

[I]t can be said that young people derive too much pleasure online from the kind of narcissism that the lifestyles of some celebrities promote. Such is a form of objectification of the human potential. People have to make mature choices. But online technology has exposed many young people to a self-centered way of life with very negative consequences. For instance, some may neglect the value and presence of people in their lives, including family and one’s commitment to society.

However, it must also be noted that modern technology as a social process is not something that is pre-determined. In contrast to the above, the internet-savvy generation today is drawn to the online world since virtual reality offers a vigorous way of life. Young people are enamored to the online world because social media creates many vibrant relationships and common ties that bind the young into groups. As such, being hooked online does not necessarily mean that an individual has become less responsible in terms of his commitments. (6)

Here-and-now risks vs. sci-fi movie ones

Hanna and Bender write:

Nevertheless, [in 2023] the nonprofit Center for AI Safety released a statement - co-signed by hundreds of industry leaders - warning of “the risk of extinction from AI,” which it asserted was akin to the threats of nuclear war and pandemics. Sam Altman, embattled CEO of Open AI, the company behind the popular language-learning model [LL] Chat-GPT, had previously alluded to such a risk in a congressional hearing, suggesting that generative AI tools could go “quite wrong.” Last summer executives from AI companies met with President Joe Biden and made several toothless voluntary commitments to curtail “the most significant sources of AI risks,” hinting at theoretical apocalyptic threats instead of emphasizing real ones. Corporate AI labs justify this kind of posturing with pseudoscientific research reports that misdirect regulatory attention to imaginary scenarios and use fearmongering terminology such as “existential risk.” [my emphasis]

But Hanna and Bender do warn about the ability to use AI to create fake texts makes the misinformation problem worse:

Unfortunately, that output can seem so plausible that without a clear indication of its synthetic origins, it becomes a noxious and insidious pollutant of our information ecosystem. Not only do we risk mistaking synthetic text for reliable information, but that noninformation reflects and amplifies the biases encoded in AI training data—in the case of large language models, every kind of bigotry found on the Internet. Moreover, the synthetic text sounds authoritative despite its lack of citation of real sources. The longer this synthetic text spill continues, the worse off we are because it gets harder to find trustworthy sources and harder to trust them when we do.

There is also risk of stolen intellectual property and old-fashioned exploitation of workers:

[AI] systems rely on enormous amounts of training data that are stolen without compensation from the artists and authors who created them. In addition, the task of labeling data to create “guardrails” intended to prevent an AI system’s most toxic output from being released is repetitive and often traumatic labor carried out by gig workers and contractors, people locked in a global race to the bottom in terms of their pay and working conditions. What is more, employers are looking to cut costs by leveraging automation, laying off people from previously stable jobs and then hiring them back as lower-paid workers to correct the output of the automated systems.

It was apparent early on in the Industrial Revolution that technological advances could have large disruptive effects on society. In fact, that was a central concern of Adam Smith’s The Wealth of Nations. So that is not new. But that also doesn’t mean that it isn’t a real problem that needs to be addressed by public economic policies. The alleged magic of the market won’t do it, unless we make an anarcho-capitalist circular argument that whatever conditions The Market produces is always the best of all possible worlds.

Hanna and Bender note that the long 2023 actors’ and writers’ strike in Hollywood was an instance of the labor movement addressing those very real problems. Market anarchy may be an ideal outcome for oligarchs, but certainly not for everyone.

They also note that much of the information on AI is put out by industry sources. So fans, users, and investors all need to take account of that limitation. They also observe that since much of the data is kept under wraps by the companies and agencies work on AI development, it is often not possible to attain independent verification of claims made, such as in one paper they cite “which claims to find “intelligence” in the output of GPT-4, one of OpenAI’s text-synthesis machines.”

I ran into a similar problem in an exercise that gave me a glimpse of that problem recently when I used a couple of LLMs (large language models) to try to find out whether there was actual evidence that large financial companies have been able to use AI to achieve improvement in productivity.

I asked the chatbots for five specific examples, and they gave me five companies: Intuit, PwC (PricewaterhouseCoopers), Xero, KPMG, and EY (Ernest & Young). Checking their websites, I found that they all do make glowing claims for AI, and they all market AI products. But their website marketing pitches that I saw were awfully short on examples of what kinds of AI tools they were offering and exactly how they would improve productivity.

When I asked my LLMs for specific independent studies on the improvements in productivity via AI at finance firms, I got what is a common quirk of LLMs right now. I got a list of specific studies with specific titles and the authors or institutions that put them together. But when I did Google searches for the specific studies, I didn’t find any that seem to actually exist. If you give an LLM a specific inquiry for sources like that, they are trained to produced results that meet the parameter of the request, i.e., a list of reports from plausible sources and even dates. But they are often not able to confirm whether the sources they name actually exist. They are trained on specific and large datasets. But at least the ones I’ve used don’t seem to have the ability to do a reality-check through the Internet on whether the sources are even real. The focus of the designs so far have tended to emphasize producing coherent and grammatically correct responses to inquiries. So the fact-checking is still up to the user.

The energy issue

Like with cryptocurrency, the amount of energy required to run "large language models" like ChatGPT is enormous.

One obvious difference between AI and mammal brains is that the mammal brains are staggeringly more energy-efficient that any AI currently on the horizon. Unless we soon are actually able to build power plants with sustainable fusion reactors, energy requirements will be a major brake on AI possibilities. Those fusion-based power plants been a constant fantasy of what we now call tech-bros since roughly the day of the Trinity a-bomb test.

The value of cryptocurrency over other means of exchanging money electronically has always been highly dubious. We can’t make the same generalization for all artificial intelligence applications. Things like the translation function on Microsoft Word are examples of AI that have been around for years. In fact, the current AI “wave” (or bubble”?) is actually the third of its kind.

Lois Parshley warns in The American Prospect that the huge amount of energy currently being used by current AI applications is something that deserves immediate attention by legislators and the public.

In early May, Google announced it would be adding artificial intelligence to its search engine. When the new feature rolled out, AI Overviews began offering summaries to the top of queries, whether you wanted them or not—and they came at an invisible cost.

Each time you search for something like “how many rocks should I eat” and Google’s AI “snapshot” tells you “at least one small rock per day,” you’re consuming approximately three watt-hours of electricity, according to Alex de Vries, the founder of Digiconomist, a research company exploring the unintended consequences of digital trends. That’s ten times the power consumption of a traditional Google search, and roughly equivalent to the amount of power used when talking for an hour on a home phone. (Remember those?) (7)

The energy issue with AI is a huge one.

Critical thinking

The fact that large established companies like Microsoft and Meta may put limits on the amount of dot-com bubble fluff and scams that come out of it. But don’t count on it.

Critical thinking about AI claims is always called for.

And since tools can often be used for a variety of purposes, AI can also be used to combat disinformation and unpack conspiracist thinking. Benjamin Radford reports:

Conspiracy theories—a perpetual and pernicious bane of skepticism and critical thinking—have traditionally been examined through several prisms, including folklore, psychology, and social psychology. More recently, the field of computational analysis has emerged to help identify, address, and mitigate rumors and misinformation. Among the field’s pioneers is Timothy Tangherlini, a professor of folklore at the University of California at Los Angeles and a fellow of the American Folklore Society. “I study narrative, how stories emerge and circulate on and across social networks; rumors, legends, conspiracy theories, and so on,” he explained in [a 2022] interview. …

Tangherlini researches narrative structures of conspiracy theories. He and colleagues note that “Despite the attention that conspiracy theories have drawn, little attention has been paid to their narrative structure, although numerous studies recognize that conspiracy theories rest on a strong narrative foundation or that there may be methods useful for classifying them according to certain narrative features such as topics or motifs” (Bandari et al. 2017). The team “developed a pipeline of interlocking computational methods to determine the generative narrative framework undergirding a knowledge domain or connecting several knowledge domains.”

Using research on public health concerns (specifically, anti-vaccination blog posts) and combining it with folklore legend research, the model consists of three primary components that populate the narrative structure. Just as every story has certain consistent elements (such as a setting, protagonist, conflict, and resolution), the model studies three main components: actants (people, places, and things), relationships between those actants, and a sequencing of those relationships. (8)

In a 2020 article by Tangherlini and others, they provide this chart of conspiracist narratives, which bears some resemblance to charts illustrating the functioning of AI neural networks: (9)

Notes:

(1) Hanna, Alex & Bender, Emily (2024): Theoretical AI Harms Are a Distraction. Scientific American 330:2 (Feb 2024), 69. <https://www.scientificamerican.com/article/we-need-to-focus-on-ais-real-harms-not-imaginary-existential-risks/> (Accessed: 2024-06-07).

(2) Artificial Intelligence Coined at Dartmouth: 1956. Dartmouth College website, n/d. <https://home.dartmouth.edu/about/artificial-intelligence-ai-coined-dartmouth> (Accessed: 2024-06-07).

(3) Parshley, Lois (2023): Artificial Intelligence Could Finally Let Us Talk with Animals. Scientific American 329:3 (Oct 2024), 46. <https://www.scientificamerican.com/article/artificial-intelligence-could-finally-let-us-talk-with-animals/> (Accessed: 2024-06-07).

(4) Kelley, Alexandra (2024): Army to seek industry help on AI. Defense One 06/27/2024. <https://www.defenseone.com/technology/2024/06/army-plans-multiple-ai-industry-partnerships/397705/> (Accessed: 2024-06-07).

5) Abraham, Yuval (2024): ‘Lavender’: The AI machine directing Israel’s bombing spree in Gaza. +972 Magazine 04/03/2024. <https://www.972mag.com/lavender-ai-israeli-army-gaza/> (Accessed: 2024-06-07).

(6) Maboloc, Christopher Ryan (2017): Social Transformation and Online Technology: Situating Herbert Marcuse in the Internet Age. Techné: Research in Philosophy and Technology 21:1, 56. Techné: Research in Philosophy and Technology 21:1.

(7) Parshley, Lois (2024): The Unknown Toll of the AI Takeover. The American Prospect 07/01/2024. <https://prospect.org/environment/2024-07-01-unknown-toll-of-ai-takeover/> (Accessed: 2024-06-07).

(8) Radford, enjamin (2024): Analyzing Conspiracies through Folklore, Epidemiology, and Artificial Intelligence. Skeptical Inquirer 48:3 (May-June 2024), 47-52. <https://skepticalinquirer.org/2024/04/analyzing-conspiracies-through-folklore-epidemiology-and-artificial-intelligence/> (Accessed: 2024-06-07).

9) Tangherlini, Timothy (2020): An automated pipeline for the discovery of conspiracy and conspiracy theory narrative frameworks: Bridgegate, Pizzagate and storytelling on the web. Plos One 06/16/2020. <https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0233879> (Accessed: 2024-06-07).

AI in National Security Decision-Making

The Carnegie Endowment has an analysis on “How AI [artificial intelligence] Might Affect Decisionmaking in a National Security Crisis.”

This brings to mind yet another strand in popular culture images of technology taking over. There was the Russian system in Stanley Kubrick’s Dr. Strangelove (1964) that would automatically launch the USSR’s. Another was WarGames (1983) starring Mathew Broderick and Ally Sheedy. At the end of Dr. Strangelove, the world pretty much comes to an end as Slim Pickens as a obsessively militaristic pilot manages to deliver a nuke onto Soviet territory, and the automatic retaliation device fires off the Soviet nukes.

By the way, that isn’t a purely fictional idea. The real-world version is (grimly) called a “dead hand” system.

WarGames had a happy ending, because the AI supercomputer system, known as Joshua, runs a quick learning routine at the instigation of its creator in which it plays about a billion games of Tic Tac Toe, after which Joshua refuses to execute the missile launch article, informing the humans that the only way to win this game is not to play. In this case, AI Joshua saves the humans from their own stupidity and recklessness, without deciding it needed to set itself up as the overlord of humanity as a result.

In a 2008 sequel (War Games: The Dead Code), an evil AI system called RIPLEY is trying to cause havoc. Fortunately for our heroes and humanity, Joshua is still around with a somewhat less-apocalyptic function of running the power grid. And he still has the moxie to block RIPLEY from his mischievous mission. Joshua had developed a somewhat dark sense of humor over the two and half ensuring decades, telling the heroes at the end: “Yes, the human race is finished. That was humor.” (Ah, but was it, though?)

Chivvis and Kavanagh report on a simulation they did:

To get a grip on how the proliferation of artificial intelligence might affect national security decisionmaking at the highest levels of government, we designed a hypothetical crisis in which China imposed a blockade on Taiwan and then convened a group of technology and regional experts to think through the opportunities and challenges that the addition of AI would bring in such a scenario.

Not surprisingly, one result was that AI could provide much more relevant information faster than less advanced systems. Information overload is a well-known problem in decision-making, in private as well as public organization. But his seems like an encouraging result from the Taiwan exercise:

[T]he experts immediately wanted to know more about the underlying AI system so that they could interpret its recommendations. They needed to understand why the system was making the recommendations it was before they could have a degree of confidence in its prescribed courses of action. They also wanted to weigh the AI’s recommendations with more traditional sources of information—specifically the actual human experts around the table. This meant that AI became just another voice in the process, one that also had to gain the confidence of the decisionmakers.  [my emphasis]

One of the most important lessons I learned about financial forecast and mathematical forecast models generally is the evaluation guideline: If a forecast looks wrong, it probably is. Meaning pretty much what was described there. People with actual experience and expertise in the field look at the forecast and ask if it makes sense based on their knowledge and experience. Because ultimately, the people responsible for the decisions are the ones who have to decide on the risks and benefits.

“AI proliferation might also slow decisionmaking by creating uncertainty about adversary intentions and forcing policymakers to ponder whether and how AI might be shaping adversary actions.” In this category, Chivvis and Kavanagh mainly discuss, the latter concern, i.e., are policymakers on the other side operating on bad information generated by AI systems? This in itself is nothing new, but keeping up with the technology that can generate misleading information is of course important.

But when I read the first part of that sentence, “AI proliferation might also slow decisionmaking by creating uncertainty about adversary intentions,” I thought, yes, US foreign-policy decisions need a lot more of that, instead off blundering around with outdated, superficial, or ill-informed assumptions.

For instance, the triumphalist rhetoric from Western leaders about Ukraine beating Russia in the current war could always be for public consumption, with those making it knowing better. But if a well-designed AI system is actually take seriously and used carefully by decision-makers, they might actually decide to modify their views. AI is not magic. But in processing large amounts of relevant information fast, it could be particularly useful in a crisis situation.

In a section with a covering-all-baes subtitle, AI Might Combat Groupthink … or Make It Worse”, they do consider the very real problem of groupthink. Irving Janis’ work on groupthink in crisis decision-making gives some useful descriptions of what it is and how it can be successfully balanced. And a well-constructed AI system could certainly be helpful in providing alternative perspectives. But the AI system will not be “Joshua” who can or will save the humans from their own decisions.

They also note, “Unfortunately, AI can also have the opposite effect of encouraging groupthink, especially in situations where the decisionmakers have high confidence—or too much confidence—in the capability of the AI system.” This strikes me as unlikely. The more common risk would be decision-makers using AI to promote their own prejudices on situations. It is likely that AI-generated military scenarios could have for some decision-makers the appearance of specificity that could be misleading. That would be an AI variation on the common problem of, if the only tool you have is a hammer, then every problem looks like a nail.

But this comment is pretty fuzzy: “Clearly [sidelining experienced experts] is a situation to be avoided, but just keeping a human in the decisionmaking loop may not be enough to prevent the AI from effectively running the show.” No, it’s the human decision-makers who are running the show. The buck stops with them, we might say. (The cryptocurrency, too, actually!)

They offer a variation on the Turing Test theme (whether people can tell whether they are interacting with a computer or a human):

In this scenario, uncertainty about the presence and role of the AI system made interpreting the intentions of the adversary very difficult for our experts. Specifically, it became unclear whether adversary moves were determined by an AI or by a human being. In an actual crisis, U.S. policymakers would probably be similarly unsure whether a human or machine is on the other side of the physical or virtual battlefield. Uncertainty about the role and presence of AI would also make signaling more difficult, increasing the risk of misperception and miscalculation and creating a perfect storm for unintended escalation even when both sides prefer to avoid conflict.

I’m not sure this is much more than a statement that evolving technology has to be used and responded to sensibly.

Chivvis and Kavanagh stress the need for international regulations and mutual understandings about AI technology, like nuclear and other technologies:

The challenge, of course, will be adopting a set of principles that all relevant parties might agree to, as well as a mechanism for ascertaining compliance. This challenge is greatly magnified by the fact that the leaders in AI innovation are commercial firms, not governments, and by the rapid speed with which AI systems are evolving and advancing. The Biden administration has sketched out a policy to guide military uses of AI and set up an AI Safety Institute to anticipate and mitigate dangerous uses of AI technology. While there is some alignment between the United States and key allies on these issues, to really have an impact, any AI arms control regime would have to include China. The two competitors held preliminary discussions about AI safety and governance in May 2024, but given strained ties and limited dialogues between Washington and Beijing, progress in the near term may remain slow. Still, U.S. policymakers should continue to push forward with willing partners where possible.

The National Security Institute has made this 2-hour video available showing a Taiwan crisis simulation decision-making process using AI:

Notes:

(1) Chivvis, Christopher S. & Kavanagh, Jennifer (2024): Carnegie Endowment 06/17/2024. <https://carnegieendowment.org/research/2024/06/artificial-intelligence-national-security-crisis?lang=en> (Accessed: 2024-19-06).

(2) Peck, Michael (2021): Dr. Strangelove: Russia's Dead Hand Nuclear System. The National Interest 09/23/2021. <https://nationalinterest.org/blog/reboot/dr-strangelove-russias-dead-hand-nuclear-system-194188>  (Accessed: 2024-19-06).

(3) WarGames:The Dead Code. Wikipedia 04/30/2023. <https://en.wikipedia.org/w/index.php?title=WarGames:_The_Dead_Code&oldid=1152523505> (Accessed: 2024-19-06).

(4) Janis, Irving. L. (1982): Groupthink: Psychological Studies of Policy Decisions and Fiascoes, 2nd edition. Dallas:Houghton Mifflin.

(5) AI vs. Human Decision-Making: Crisis in the Taiwan Straits Wargame. National Security Institute YouTube channel 02/07/2024. <https://youtu.be/dk1nryglbTg?si=apueNcpG-9ge2O9y> (Accessed: 2024-19-06).

Artificial Intelligence (AI): reality and hype

Christiane Amanpour recently presented this reality-based discussion of artificial intelligence (AI) with MIT economist David Autor that is cautiously optimistic and informative about the subject. (1)




Parsing the hype

One guideline I use for myself in thinking about AI is a Not-Skynet-or-Mister-Data measure. If we think of AI as an android that can pass the Turing Test, i.e., humans using the system are unable to distinguish whether they are talking to a machine or a human being, then AI is still a fantasy. A well-grounded science-fiction fantasy maybe, but still a fantasy.

In the last decade, “critics have continually noted that the Turing Test is less about building an intelligent system as it is about building one that deceives people most effectively.” Being clever at mimicking human behavior is not the only element of intelligence for AI machines. Deception and intelligence are not identical, in other words. (2)

But neural networks that facilitate processing and use of data are real systems with real capabilities. They can evaluate large numbers of data points and make numerical calculations and derive probabilities faster than human mathematicians. And machines today like mobile phones can process the same amounts of data that decades ago would have required huge physical computers.

Yet, as Cristoph Kehl has observed, “When - and if in any case - a strong AI is realizable which comes close to human intelligence is a question over which we can only speculate.” (3)

Kehl takes the application of AI to elder care as a paradigmatic case of how to think about the possibilities, uses, practicality, and ethics of AI.

The idea that the rising need for elder care can be satisfied by Mister Data-type service androids is blind optimism, though it’s a favorite fantasy for AI boosters. For wealthy countries, the much more feasible and readily possible solution lies in immigration and sensible laws and educational practices to go with it.

As Kehl explains, this is a good field in which to conceptualize the barriers between humans and AI machines: “Because the role of neurotechnology is mostly limited to the therapeutic field, service robotics for geriatric care is the driving force behind the dynamics of the dissolution of boundaries.” (4)

That also makes the field a good way to conceive of practical applications and their limits. Robots can delivered meals to rooms in a nursing-care facility. But the field also requires the emotional interaction other humans provide and the kind of nuanced evaluations of behavior that medical samples and AI calculations cannot provide.

Kehl also notes, “The large amount of energy AI requires is also part of the differentiation between humans and machines.”

AI does require a lot of energy, which also proved to be a more limiting factor in cryptocurrency. While AI models adapt knowledge from human brain operations, they don’t come remotely close to the human brain’s energy efficiency at this point. (5)

[A]rtificial systems do not yet actually have consciousness and free will. Consciousness in the sense of subjective experience would be required, for example, in order to be able to feel moral emotions such as compassion or guilt. Free will opens up the possibility of deciding against an option for action that is recognized as moral and of [instead] acting immorally. Artificial systems do not yet have this ability either - and should not have it, in order to protect the user. (6)

AI Waves of development and AI “Winters”

One thing to note in the current phase of AI excitement is that a large part of the hyperventilating about AI is actually saying, there were huge technological developments in the past that transformed the world and AI could do that, too! It could, and it is doing so to some extent.

But in those kinds of presentations, you may have to look very hard to get information about what the actual promising developments are and why they are promising. A recent example I encountered was a German book titled, Droht das Ende der Experten? ChatGPT und die Zukunft der Wissensarbeit (Is the End of Experts Impending? (ChatGPT and the Future of Knowledge Work). (7) But just because the cotton gin or the steam engine hat dramatic effects on the world economy, doesn’t mean that every novel development will have the same fate. Framing AI in that way is more hype than analysis.

David Autor’s interview above is an example of a meaningful Long View picture of technology developments that’s not just an instance of superficial promotional speculation. It includes a useful framework of comparing the jobs of air traffic controller and crossing guards that illustrates why some things can be more easily automated (or AI-ed) than others.

Manuela Lenzen reminds us that there have been two previous waves of AI research and development that were followed by what she calls “AI winters.” (7) The current Wikipedia article on the topic notes, “There were two major winters approximately 1974–1980 and 1987–2000,” though it notes that different sources date the two slightly differently. (8) It designates the early 2020s as a third AI “spring.” When ChatGPT in November 2022 was made publicly available online, that was an important moment in raising general public awareness of the current version of artificial intelligence.

David Autor’s interview above is an example of a meaningful Long View picture of technology developments that’s not just an instance of superficial promotional speculation. It includes a useful framework of comparing the jobs of air traffic controller and crossing guards that illustrates why some things can be more easily automated (or AI-ed) than others.

Manuela Lenzen reminds us that there have been two previous waves of AI research and development that were followed by what she calls “AI winters.” (7) The current Wikipedia article on the topic notes, “There were two major winters approximately 1974–1980 and 1987–2000,” though it notes that different sources date the two slightly differently. (8) It designates the early 2020s as a third AI “spring.” When ChatGPT in November 2022 was made publicly available online, that was an important moment in raising general public awareness of the current version of artificial intelligence.

As MIT Technology Review put it in 2023:

We’ve reached peak ChatGPT. Released at the end of November as a web app by the San Francisco–based firm OpenAI, the chatbot exploded into the mainstream almost overnight. According to some estimates, it is the fastest-growing internet service ever, reaching 100 million users in January, just two months after launch.

Through OpenAI’s $10 billion deal with Microsoft, the tech is now being built into Office software and the Bing search engine. Stung into action by its newly awakened onetime rival in the battle for search, Google is fast-tracking the rollout of its own chatbot, based on its large language model PaLM. Even my family WhatsApp is filled with ChatGPT chat. (9)

Since the current AI wave is functioning as the tech industry’s Next Big Thing at the moment, it’s worth remembering that the term “artificial intelligence” made its way into the general human vocabulary through being popularized by the Dartmouth Summer Research Project on Artificial Intelligence (DSRPAI) in 1956, which was hosted by John McCarthy and Marvin Minsky, and featured the presentation of the Logic Theorist program and which was “designed to mimic the problem solving skills of a human and was funded by Research and Development (RAND) Corporation. It’s considered by many to be the first artificial intelligence program.” (10)

Notes:

(1) AI Could Actually Help Rebuild the Middle Class, Says MIT Economist. Democracy Now! YouTube channel 04/09/2024. <https://youtu.be/1Epl0bI_jW4?si=GziVSWjnHDIOuvpJ> (Accessed: 2024-10-04).

(2) Lenzen, Manuela (2023): Künstliche Intelligenz. Was sie kann & was uns erwartet (4.aktualisierte Auflage), 29. München: C.H. Beck. My translation from the German.

(3) Kehl, Christoph (2018): Entgrenzungen Zwischen Mensch und Maschine, oder Können Roboter zu Guter Pflege Beitragen?): Aus Politik und Zeitgeschichte (APuZ) 6-8. My translation from the German. “Wann – und ob überhaupt – eine starke KI realisierbar ist, die der menschlichen Intelligenz gleichkommt, ist eine Frage, über die sich nach wie vor nur spekulieren lässt.”

(4) Kehl, Christoph, Ibid. My translation from the German.

(5) Whitten, Allison (2024): Ein Chip manch dem Vorbild Des Gehirns. Spektrum SPEZIAL Biologie Medizin Hirnforschung 1.24, 62-65.

(6) Misselhorn, Catrin (2018): Mascinenethik und "Artificial Morality": Können un Sollen Maschines Moralisch Handeln? Aus Politik und Zeitgeschichte (APuZ) 6-8, 29-33. My translation from the German.

(7) Holtel, Stefan (2024): München: Verlag Franz Vahlen.

(8) Lenzen, Manuela (2024): Künstliche Intelligenz.Fakten, Chancen, Risiken. München: C.H. Beck.

(9) AI winter. Wikipedia 04/29/2024. <https://en.wikipedia.org/w/index.php?title=AI_winter&oldid=1221339673> (Accessed: 2024-03-05).

(10) Heaven, Will Douglas (2023): ChatGPT is everywhere. Here’s where it came from. MIT Technology Review 02/08/2023. > (Accessed: 2024-03-05).

(11) Anyaha, Rockwello (2017): The History of Artificial Intelligence. Harvard.edu 08/28/2017. (Accessed: 2024-03-05).

Artificial Intelligence - what is it? (Besides investor hype, of course!): The Turing Test and neural networks

The main booster-hype around cryptocurrency seems blessedly to have died down. Who knows when it may pop up again? As he put it:

To the extent that Bitcoin is the “future of money,” then, it is only the future of money in situations of extreme crisis or deprivation—I suspect a lot of the pro-crypto people who understand its present-day uselessness are betting on a future collapse of the global economic system, although I think they overestimate the chances that Bitcoin itself could keep functioning effectively in such a nightmare scenario (someone has to maintain the actual wires). (1)

This is not to say that cryptocurrency has no use - alongside its various problems. Creating an anarcho-libertarian market paradise is not one of them. Large corporations could use it to create a set of competing private currencies. The European Central Bank (ECB) is working on setting up an official crypto version of the euro. (2) This could be useful in blocking a corporate-oligarchical system of parallel currencies. It could also potentially be used to give the central bank an additional tool to limit the capability of the private banking system to cause financial crises or exacerbate them.

Of course, an official currency issued by a central bank for the purpose of stabilizing the economy and preventing damage to the system by wealthy private actors is about as far from the anarcho-libertarian ideal hyped by the crypto-bros as we could get!

Now artificial intelligence (AI) is the topic of a new round of investment hype for companies looking to become the leader of The Next Big Thing.

So it’s worth taking some time and effort to understand what AI really is while also watching who may become the Google or Apple of AI. Which could be Google or Apple, of course.

Franz Schneider offers this definition of AI:

To what extent could the performance of a machine be judged as ‘intelligent’ – that is, commensurable (measurable in the same terms) with human intelligence? Since the Turing test, machines have been judged as ‘intelligent’ by comparing their behavior with social conventions. Cybernetics investigated this question in a different way, that is, by postulating a common ‘mechanism’ (whatever logical or physiological) between humans and machines. But in the decades prior to cybernetics and computer science, psychometrics had already turned human intelligence into a quantifiable (and potentially computable) object. In the early twentieth century, [Charles] Spearman, for instance, proposed the statistical measurement of ‘general intelligence’ (or g factor) as the correlation between unrelated tasks in a skill test. For Spearman, these correlations mathematically demonstrated the existence of an underlying cognitive faculty that common sense would refer to as ‘intelligence’. (3) [my emphasis]

Two important concepts in the development of AI are the Turing Test and neural networks.

The Turing Test

Part of the idea of “artificial intelligence” is that it is more advanced that simple mathematical calculations. Interactivity between humans and the AI device is one of the ideas associated with it. The “Turing test” was a concept developed by the British mathematician Alan Turing (1912-1954) that was meant to show whether a device met that threshold.

In a Turing Test … a person poses questions to other people and to a computer. The person and the computer answer in chat format (without visual or audio contact). The Turing Test is passed [by the device] when the person posing the questions cannot say which of his “conversation partners” is a machine. (4)

Neural networks

The use of neural networks by the machines was also part of the concept of AI.

For decades, neuroscientists’ theories about how brains learn were guided primarily by a rule introduced in 1949 by the Canadian psychologist Donald Hebb, which is often paraphrased as “Neurons that fire together, wire together.” That is, the more correlated the activity of adjacent neurons, the stronger the synaptic connections between them. This principle, with some modifications, was successful at explaining certain limited types of learning and visual classification tasks.

But it worked far less well for large networks of neurons that had to learn from mistakes; there was no directly targeted way for neurons deep within the network to learn about discovered errors, update themselves and make fewer mistakes. “The Hebbian rule is a very narrow, particular and not very sensitive way of using error information,” said Daniel Yamins, a computational neuroscientist and computer scientist at Stanford University. (5)

Improvements on AI calculations continue to draw on the knowledge of how human intelligence functions. That may sound banal, because we use AI to work with humans. But the frameworks actually used are important to keep in mind. And also the fact that in the current state of development, AI is far away from the formation of intelligence of which the human brain is capable. Mr. Data and Skynet may be in our future. But they aren’t here yet.

TensorFlow.org has an interactive chart “playground” (6) to illustrate how neural networks function:

Notes:

(1) Robinson, Nathan J. (2021): Why Cryptocurrency Is A Giant Fraud. Current Affairs 04/22/2021. <https://www.currentaffairs.org/2021/04/why-cryptocurrency-is-a-giant-fraud> (Accessed: 2024-24-03).

(2) Schneider, Franz (2024): Digitaler Euro – Die Geister, die man ruft. Makroskop 15.03.2024. <https://makroskop.eu/09-2024/digitaler-euro-die-geister-die-man-ruft/> (Accessed: 2024-24-03).

(3) Pasquinelli, Matteo (2023): The Eye of the Master: A Social History of Artificial Intelligence (ebook). London & New York: Verso.

(4) Range, Thomas (2018): Mensch Fragt, Mascine Antwortet.Wie Künstliche Intelligenz Wirtschaft, Arbeit und unser Leben verändert. Aus Politik und Zeitgeschicte 68:6-8 , 16. Translation from the German is mine.

(5) Ananthaswamy, Anit (2024): Programm mit Köpfchen. Spektrum Spezial BMH 1.24, 52. English quote is from the original unsigned and undated article: Artificial Neural Nets Finally Yield Clues to How Brains Learn. Quanta Magazine. <https://www.quantamagazine.org/artificial-neural-nets-finally-yield-clues-to-how-brains-learn-20210218#> (Accessed: 2024-24-03).

(6) https://playground.tensorflow.org/ (Accessed: 2024-24-03).