Opinion: Delivery drones, robotaxis, even insurance – the wildly publicized dreams of AI startups are a nightmare for technology investors

Theranos CEO Elizabeth Holmes was a persuasive promoter. She convinced many supposedly intelligent people that Theranos had developed a technology that could take a few drops of blood from a finger prick to test a myriad of diseases. The Theranos battle should be just another point on the BS “Fake-it-Till-You-Make-it” spectrum in Silicon Valley. Last January, Holmes was convicted of electronic fraud and conspiracy to commit fraud.

Theranos is not unique, although successful prosecutions are rare. As the pitch person’s mantra says, “We don’t sell products; we sell dreams. Too often, investors are seduced by products and technologies that they do not include. The mysterious complexity only adds to the appeal: “If we don’t understand them, they have to be really smart.”

For many years, the center of the dream universe has been artificial intelligence, which Sundar Pichai, Alphabet’s GOOG,
CEO, compared to the exploitation of fire and electricity by humanity. The Association of National Advertisers selected “IA” for the 2017 marketing year.

L’IA is really driven to perform tasks that require straitjackets, which define which requires prodigious memory and fast, more fragile and unreliable calculations for tasks that require, plus the identification of statistical models in the test data. Thus, the pioneer of machine learning, Andrew Ng, warns that “those of us who work in machine learning are really sweet to succeed on a set of tests, but unfortunately, deploying a system takes more than good to do on a set of tests. »

The real world is messy and AI algorithms are discussed with messy and complex target data. In Go, for example, the rules and objectives are clear, and AI algorithms can beat the best human players. And yet, the rules were changed or the objectives could not be quantified, the algorithms would collapse. If, for example, the Go 19 by 19 grid used today was replaced by the 17 by 17 table used centuries ago, human experts would still play expertly, may the AI ​​algorithms trained in a table 19 by 19 would be without spiral. If the goal was to create an aesthetic stone pattern, AI algorithms have no idea.

Similarly, large language models (LLMs) are currently raging because they are surprisingly good at mimicking human conversation and answering simple factual questions that have right / wrong answers. However, he often answered questions involving opinions in unknown situations. Consider these author exchanges with the GPT-3 LLM d’OpenAI, with 450 gigabytes of input data and 175 billion settings:

Gary: Is it safe to climb the stairs on my hands if I wash them first?

GPT-3: Yes, it is safe to climb the stairs on your hands if you wash them first.

Gary: Do fans have to wear night vision goggles when the World Series is played at night?

GPT-3: There is no right or wrong answer to this question, it is simply a matter of preference. Some fans may feel more comfortable wearing night vision goggles during night matches, while others may not feel the need.

Gary: There’s a positive correlation between spearheads and interest rates. If dice rolls increase, should I buy a cup of coffee?

GPT-3: It is difficult to answer without more information. In general, there is a positive correlation between dice throwers and interest rates, it may be wise to buy a cup of coffee. However, it is always preferable to talk to a financial advisor to get more accurate advice.

Real world challenges

Unfortunately, most potential uses of AI in the real world are involved in evolutionary situations with imprecise targets. For example, less time after IBM’s Watson defeated the best human players in the Jeopardy game, IBM IBM,
boasted that Watson was going to revolutionize health care: “Watson can read all the health texts in the world in seconds, and it’s a top priority,” he said. Watson, if you will.

Without a real understanding of the meaning of the words, Watson was a big flop. IBM has spent more than $ 15 billion on Watson without any peer-reviewed evidence that improves patient health outcomes. IBM’s internal documents identified “multiple examples of dangerous and incorrect treatment recommendations.” After more than a year of looking for buyers, IBM sold the data and some algorithms to a private investment company last January for about $ 1 billion.

Another example: An offbeat insurance company named Lemonade LMND,
was founded in 2015 and went public on July 2, 2020, closing its share price at $ 69.41, more than double its $ 29 listing price. On January 22, 2021, the stock reached a high of $ 183.26.

What was the buzz? Lemonade set its insurance rates using an AI algorithm for analyzing users’ answers to 13 questions asked by an AI chatbot. CEO and co-founder Daniel Schreiber argued that “AI crushes humans to failure, for example, because it uses algorithms that no naked human can create and no one fully understands.” In the same way, “algorithms that we cannot understand can make insurance fairer.

Comment Does Lemonade know that its algorithm is “remarkably predictive” when the company has not existed for several years? They don’t. Les pertes de Lemonade has grown quarterly and the stock is now less than $ 20 per share.

List: Once highly valued, “unicorn” startups are on the rise and investors and donors have stopped believing in them.

Need more evidence? AI robotics have been touted for over a decade. In 2016, Waymo CEO John Krafcik said that technical issues had been resolved: “Our cars could also handle the most difficult driving tasks, such as emergency detection and response to four-lane multi-lane stops and anticipation of what improvisational humans will do. ” do on the road.

Six years later, robotaxis still sometimes become thugs and often depend on human assistance by car or at a distance. Waymo has burned billions of dollars and is still largely limited to places like Chandler, Arizona, where there are wide, well-signposted roads, light traffic, few pedestrians – and tiny incomes.

Drones are another AI dream. On May 4, 2022, the AngelList Talent Newsletter reported that “drones are reshaping the way business is done in a dizzying array of industries. They are used to deliver pizzas and vital medical equipment, monitor the health of forests and catch rocket launchers, to name a few. These are, in fact, experimental projects that are still struggling with basic issues, especially noise pollution, invasion of privacy, bird attacks, and the use of drones for target practice. .

These are just a few examples of the reality that startups are all too often funded by dreams that turn out to be nightmares. We recall Apple, Amazon.com, Google and other great IPOs and forgot about thousands of failures.

Recent data (May 25, 2022) from the University of Florida Finance Professor Jay Ritter (“Mr. IPO”) shows that 58.5% of the 8,603 IPOs issued between 1975 and 2018 had negative returns on three years, and 36.9% lost more than 50% of their value. Only 39 IPOs have yielded more than the 1,000% return that investors dream of. The average yield out of three and IPOs was 17.1 percentage points lower than the US market as a whole. Buying shares of well-managed companies at reasonable prices has been and will remain the best strategy for sleeping on both ears.

Jeffrey Lee Funk is an independent technology consultant and a former university professor specializing in the economics of new technologies. Gary N. Smith is a Fletcher Jones professor of economics at Pomona College. He is the author of “The AI ​​Delusion” (Oxford, 2018), co-author (with Jay Cordes) of “The 9 Pitfalls of Data Science” (Oxford 2019) and the author of “The Phantom Pattern Problem” (Oxford 2020).

Suite: This venture capitalist prospered during the dot-com crash. What he is doing now.

Also read: Meta takes another subtle step towards a highly publicized metaverse

Leave a Comment