{"id":2202,"date":"2024-06-22T07:39:38","date_gmt":"2024-06-22T11:39:38","guid":{"rendered":"https:\/\/bhide.net\/wordpress_files\/?p=2202"},"modified":"2024-08-02T07:42:05","modified_gmt":"2024-08-02T11:42:05","slug":"more-skeptical-remarks-about-ai","status":"publish","type":"post","link":"https:\/\/bhide.net\/wordpress_files\/index.php\/more-skeptical-remarks-about-ai\/","title":{"rendered":"(More) Skeptical Remarks about AI"},"content":{"rendered":"\n

I made the remarks below to a CEO forum on June 21 2024. Generally the AI enthusiasts were over the top vocal. Skeptics were quiet but quietly supportive of my viewpoint<\/em><\/h3>\n\n\n\n

Suppose someone said that smartphones were on the cusp of generating widespread transformations.<\/p>\n\n\n\n

You might reasonably ask, \u201cWhere have you been these last twenty years, Rip Van Winkle?\u201d<\/p>\n\n\n\n

Smartphone apps like Uber and Airbnb have revolutionized transport and travel. Mobile search and social media have crushed mainstream media and advertising.<\/p>\n\n\n\n

Given how far we have already come, is it likely that smartphones are at an inflection point? Similarly, with AI. Its applications have already been transformational<\/a>. Indeed, it is AI tools and techniques that make smartphones smart. Nearly every smartphone app \u2013 from texting to sexting, mapping to matchmaking, video editing to streaming, Uber ridesharing to Airbnb rentals \u2013 incorporates AI. When we speak to our phones asking for weather forecasts or driving directions, we engage AI\u2019s Natural Language Processing capabilities.<\/p>\n\n\n\n

Moreover, AI\u2019s widespread use precedes and goes far beyond smartphones. A 1956 workshop at Dartmouth kicked off academic AI research. In the following decades, practical applications evolved. Starting in the 1970s, George Lucas\u2019s Star Wars epics dazzled audiences with AI special effects and animations. \u2018Fuzzy logic\u2019 proposed by UC Berkeley\u2019s AI guru, Lotfi Zadeh, in 1965, was used to control a Japanese subway in 1987. By 1990, Japanese consumer electronics companies were using fuzzy logic in camcorders, vacuum cleaners, room heaters, and air-conditioners.<\/p>\n\n\n\n

In 2006 \u2013 a year before Apple\u2019s iPhone \u2013 Oxford\u2019s Nick Bostrom noted<\/a> that cutting-edge AI had \u201cfiltered into general applications, often without being called AI because once something becomes useful enough and common enough it\u2019s not labelled AI anymore.\u201d<\/p>\n\n\n\n

Sixteen years later, the claim that AI has just reached a take-off stage. is perplexing. Merely maintaining historical growth rates from a high base should be a challenge.<\/p>\n\n\n\n

Looking more closely at how AI became mainstream is instructive.<\/p>\n\n\n\n

Traditional pre-AI software applications performed deterministic calculations. Payroll processing and optimizing complex operations were archetypal applications.<\/p>\n\n\n\n

More often than not, however, uncertainties frustrate demonstrably correct solutions. Ambiguous information or incomplete knowledge makes calculating what\u2019s truly best impossible. We must make do with guesses and approximations. Likewise, we often don\u2019t use numbers or algebraic symbols to specify problems or discuss solutions. From everyday speech to Supreme Court deliberations, our discourse relies on ambiguous language \u2013 including analogies and metaphors.<\/p>\n\n\n\n

Lofti\u2019s 1965 \u201cfuzzy logic\u201d and natural language programming thus epitomize the more realistic aspirations of AI.<\/p>\n\n\n\n

But how to combine the digital computer\u2019s capacity to flawlessly manipulate 1s and 0s with the incompleteness and imprecision of human knowledge and discourse?<\/p>\n\n\n\n

One early approach incorporated specialized expertise. Medical rules of thumb were a popular basis for the early expert systems. But, this approach was limited to problems where experts had codifiable knowledge.<\/p>\n\n\n\n

 Other AI applications that used statistical approximations. Humans merely specified the data \u2013 text, images, not just numbers — from which computers inferred statistical patterns. No understanding of the underlying process or consideration of contextual meaning was necessary. The dictum, repeated endlessly in elementary statistics classes, that \u201ccorrelation is not cause\u201d was brushed aside. AI programs did not even have to be told which variables mattered or to what degree. They used data mining to calculate variable weights that best fit the observations.<\/p>\n\n\n\n

AI programs used statistical correlations to mimic natural language. Actual natural language often requires reading minds — contextual interpretation of intent. The meaning of a simple \u2018what!\u2019 depends on context and tone. Going back to MIT\u2019s Eliza, a 1960s-era psychotherapeutic chatterbot, AI programs used correlations as substitutes for any mindreading.<\/p>\n\n\n\n

Statistical AI could also improve through trial and error. But again, this \u2018machine learning\u2019 did not require domain expertise, judgments about \u201clessons learned,\u201d or understanding or consideration of context.<\/p>\n\n\n\n

Nonetheless, the cost-effectiveness of statistical AI that did not require specialized expertise vastly broadened the scope of AI applications. Google\u2019s search algorithm, which handily outperformed Yahoo\u2019s human catalogers of the internet, was a striking example.<\/p>\n\n\n\n

At the same time, AI hasn\u2019t sailed smoothly in every sea. Belying dire predictions, AI did not dominate or displace human\u2019 knowledge work.\u2019 Knowledge-intensive jobs grew, and wages stayed high.<\/p>\n\n\n\n

AI even failed to automate many tasks that don\u2019t require much thinking or training. Going back to Apple\u2019s much-ridiculed 1993 Newton, handwriting was supposed to replace typing. In 2001, Bill Gates predicted that pen-based tablets would become \u201cthe most popular form of PC sold in America\u201d<\/a> in five years. They didn\u2019t come close. Now, finally, convertible PCs with pens and touch screens have found a market, but keyboards remain the dominant input device. AI-enabled handwriting and voice recognition remain frustratingly hit or miss. Similarly, we usually still prefer the precision and accuracy of clicking or tapping on a button to giving voice instructions to personal assistants (like Siri or Alexa).<\/p>\n\n\n\n

Where has the accuracy of statistical AI been acceptable, and where has it not?<\/p>\n\n\n\n

Accuracy often depends on the ambiguity of inputs and outputs. Printed words that use standard fonts are less ambiguous than idiosyncratically handwritten words. Unsurprisingly, Optical Character Recognition software scans printed books and documents far more accurately than handwriting recognition programs.<\/p>\n\n\n\n

Ambiguous outputs similarly undermine machine learning. Unquestionably correct or wrong results have helped make face recognition highly accurate. In contrast, correctly deciphering spoken words (\u201cthere\u201d or \u201ctheir\u201d?) requires knowing the speaker\u2019s intent. But, statistical correlations cannot reliably discover intent just as they cannot establish cause.<\/p>\n\n\n\n

Accuracy also depends on the stability and uniformity of the process that generates the data used by AI applications. Physical or physiological processes, governed by invariant laws of nature, are usually stable. In contrast, human behavior and choices are subject to the whimsical vagaries of social attitudes and the zeitgeist. Statistical predictions about creditworthiness or purchasing behavior can, therefore, be highly inaccurate.<\/p>\n\n\n\n

Data produced by a uniform process provides a more reliable basis for statistical inference. For example, OCR algorithms scan text more accurately if trained with materials in the same language and script. Conversely, data shaped in diverse ways by different contextual factors \u2013 if the observations are likeunhappy families unhappy in their own way \u2013 can make statistical inferences practically useless.<\/p>\n\n\n\n

Acceptable accuracy depends on the cost of mistakes — the stakes — and the price-performance of the alternatives. Nearly every ad that Google and Meta Platforms throw at me is utterly remote from my interests. But the stakes are low and even the wildly inaccurate targeting of algorithmic advertising beats the alternative of blind advertising.<\/p>\n\n\n\n

In some creative applications of AI accuracy can be both unknowable and irrelevant. There are no correct special effects in Star Wars movies or animations in video games and cartoons. There is no objective benchmark for restoring old movie prints — who knows what the original looked like? But, automated AI restoration wins because it is much cheaper and faster than human restoration.<\/p>\n\n\n\n

Turning to the current AI mania.<\/p>\n\n\n\n

Ignorance of AI\u2019s seven-decade history may explain some over-the-top predictions about its future. But even some savvy techies who are aware of what came before assert that Large Language Models \u2013 often now conflated with all of AI \u2013 are game changers. A veteran software entrepreneur believes AI is still in its \u201cearly infancy.\u201d He argues that \u201cearlier incarnations, such as protein folding and chess playing, were esoteric and of little relevance to the general public. The chat interface to LLMs has suddenly made AI accessible to the wider public. New ideas and applications are exploding. The real creativity is coming from people using it and suggesting new uses, rather than from the engineers creating it.\u201d<\/p>\n\n\n\n

 I believe it is fair to say that before LLMs, most people were passive consumers, often unaware of the AI in their mobile phones, search engines, and social media. Certainly, LLMs have an arresting capacity for seemingly intelligent, natural language conversations with non-technical users, and they offer to automate several analytical and creative tasks. Could these abilities make LLMs a \u201ckiller app\u201d for AI to an even greater degree than the AI that has long been embedded in smartphones?<\/p>\n\n\n\n

The analogy with spreadsheets is seductive. Spreadsheets had simple user interfaces that allowed people with limited technical expertise to build useful programs. Running on cheap personal computers, they offered compelling value in many applications that did not require the power of mainframes. Symbiotically, they helped expand the personal computer market, prompting investments in better computers.<\/p>\n\n\n\n

LLMs have even simpler and more natural user interfaces than spreadsheets. Yet underneath their hoods, LLMs run statistical engines with the same statistical issues that delineated the practical scope of earlier AI applications. As with earlier AI, LLMs can shine in creative applications, such as image generation, where accuracy is irrelevant. Conversely, as with other statistical AI models, ambiguous inputs and outcomes derail their reliability and limit self-corrective learning. They can trip over data that is not generated by a stable process or is highly dependent on context.<\/p>\n\n\n\n

Relying on statistical correlations rather than deductive logic or math, LLMs have offered bizarre solutions to reasoning problems<\/a>, highlighting, for example, the risks of being attacked by a cabbage while rowing across a river. The Khan Academy\u2019s AI tutor for kids, struggles with elementary math.<\/a> (It miscalculated subtraction problems such as 343 minus 17, couldn\u2019t consistently round answers or calculate square roots, and typically didn\u2019t correct mistakes when asked to double-check its solutions.)<\/p>\n\n\n\n

Throwing every possible kind of data into LLMs\u2019 training pots does not improve accuracy and reliability. Medical data does not make responses to legal or engineering questions any better. Training on Swahili literature does not sharpen statistical summaries of Shakespeare\u2019s plays. Bulking up LLMs with disparate data so that LLMs can answer every question under the sun may increase their propensity to fantasize or hallucinate.<\/p>\n\n\n\n

Spreadsheets, in contrast, didn\u2019t overpromise and underdeliver. They didn\u2019t tell jokes or write essays, but for their more targeted functions, they followed the user\u2019s instructions precisely and correctly.<\/p>\n\n\n\n

The chatty user-friendliness of LLMs isn\u2019t a free lunch. It may well be a significant limitation. Yes, users need less knowledge of input rules and conventions than in their interactions with a spreadsheet, traditional search engine, or photo editor. But free-form inputs are also more ambiguous. Natural language prompts are more likely to evoke inaccurate or useless responses than traditional keyword searches.<\/p>\n\n\n\n

In low-risk uses people will tolerate LLM mistakes for convenience as they do with autocomplete howlers in their text messages. The multi-trillion-dollar question is whether the benefits from low-stakes uses can  cover the costs.<\/p>\n\n\n\n

One important reason for the nearly immediate popularity of spreadsheets (besides their ease of use) was that they ran on personal computers and not expensive mainframes. Similarly, Uber and Airbnb apps provided cheap, reliable alternatives to taxis and hotels through smartphones that users already owned. In contrast, LLMs require users to purchase more expensive hardware. Moreover, user hardware accounts for a fraction of the costs of building, training, and operating LLMs. For now, and as in the 1999 internet bubble, manic investors are willing to subsidize uneconomic uses. What happens when the music stops?<\/p>\n\n\n\n

At best, LLMs are akin to a new high-powered automobile engine that can win car races but makes too much noise and guzzles too much gas for street use. The hype notwithstanding, LLMs aren\u2019t like Nikola Tesla\u2019s alternating current inventions that drastically changed the economics of electrification.<\/a> Why then gamble on the transformative acceleration of AI and ignore so many other possibilities for innovation and operational improvements the world offers?<\/p>\n","protected":false},"excerpt":{"rendered":"

I made the remarks below to a CEO forum on June 21 2024. Generally the AI enthusiasts were over the<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":true,"template":"","format":"standard","meta":{"colormag_page_container_layout":"default_layout","colormag_page_sidebar_layout":"default_layout","footnotes":"","_links_to":"","_links_to_target":""},"categories":[4,33,35],"tags":[40,13],"class_list":["post-2202","post","type-post","status-publish","format-standard","hentry","category-innovation","category-opeds-and-media","category-ruminations","tag-artificial-intelligence","tag-innovation"],"_links":{"self":[{"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/posts\/2202","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/comments?post=2202"}],"version-history":[{"count":2,"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/posts\/2202\/revisions"}],"predecessor-version":[{"id":2204,"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/posts\/2202\/revisions\/2204"}],"wp:attachment":[{"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/media?parent=2202"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/categories?post=2202"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/bhide.net\/wordpress_files\/index.php\/wp-json\/wp\/v2\/tags?post=2202"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}