Is artificial intelligence a bubble?

In the early days of January, François Chollet, a French AI expert working for Google, publicly discussed how these weeks’ rhetoric about AI has evoked what has affected the cryptocurrency sector in recent years. The current climate around artificial intelligence has a lot of similarities to it Web 3 In 2021 to make me uncomfortable, ”he did written on Twitter, stating that many people expect AI to have a “civilization-altering effect within two to three years.” Although there is no certainty about it.

Chollet’s opinion caused a lot of discussion because it did not come from a figure with little affinity for or against the development of the sector but from an expert in deep learning (or Deep learning, a field of study closely related to artificial intelligence, in which the program is exposed to large amounts of data, so that it can learn to perform certain tasks). Moreover, the engineer warned his research environment against the logic of tech hype — that is, from the hyper-and intermittent enthusiasm for novelties — that in recent years has affected Web3, a term that includes cryptocurrencies, NFTs And Metaverse. Indeed, in recent years, many companies, entrepreneurs, and investors have focused heavily on these sectors, presenting them as the new horizons of the Web and beyond.

However, since last spring, the cryptocurrency sector has begun to show signs of a deep crisis. Since then it has gone through a period of scandals, from the collapse of Terra Luna (a certain cryptocurrency whose value should be more stable, in this case pegged to the value of the US dollar) to that ftxDuring which the value of bitcoin, the most widely used and widely used cryptocurrency, fell from 56 thousand euros in November 2021 to 21 thousand euros today, in what industry experts call “encryption winterThe sector’s winter.

Given the similarities between encryption and artificial intelligence, there are some who fear that the great enthusiasm generated by products such as gpt chatAnd give herAnd MidJourney AI and Stable Diffusion – the main artificial intelligences with which text, images and videos can be generated – can lead to a conclusion similar to that of cryptocurrency. last year, according to one Appreciation affiliate The New York TimesThe artificial intelligence sector recorded 78 large investment deals totaling at least $1.3 billion. This year promises to be even more active: just in the past few days, Microsoft has formalized an investment of about ten billion dollars in OpenAI, the company that develops ChatGPT, which is already cooperating with Microsoft.

Although the volume of trading and the interest generated are similar in some respects, many observers believe that the phenomenon of artificial intelligence is fundamentally different from cryptocurrency. How do He made it clear Derek Thompson, newspaper reporterAtlantic, “cryptography was money without any use” while services like ChatGPT are the opposite, or “useful tools without money, at least for now”. As we’ve seen, things change quickly on this front as well, but the differences persist.

According to Thompson – and many other experts in the field – the term “artificial intelligence” is often used incorrectly or to refer to a component of a very large sector. The technology that at the moment turns out to be revolutionary is generative technology, that is, tools capable of creating new content, but artificial intelligence is also something else and we have been living with it for some time now: when we use Instagram or YouTube, the order in which we see the contents in our feed is determined by Artificial intelligence. In this sense, Thompson explained, “AI systems have already proven themselves in a way that cryptography has never succeeded.”

The complex and unintuitive nature of blockchain, the technology upon which cryptocurrencies and NFTs are based, has not helped his assertion. Most of the discussions regarding its applications have been based on distant technologies or scenarios, which would have required an unlikely mass adoption of the blockchain itself to materialize. Although it is often presented as a still immature technology that can be improved, in fact, the invention of the blockchain dates back to 2008, by the anonymous Satoshi Nakamoto. Fifteen years have passed since then, the same years that separate us from the sale of the first iPhone (the difference is only a few months).

This is enough time for a truly revolutionary product to establish itself, conquer the market and change social, cultural and economic habits. However, in all these years, the most ardent blockchain prophets have failed to propose a convincing, useful, and exclusive application for a technology that has not been linked to mere speculation. Early in 2020, BBC king Note how blockchain has “struggled to find a job outside of the cryptocurrency business”; last year, Wired he had Known Even “useless”.

In this light, it is clear that generative AI such as ChatGPT, although it did not appear out of nowhere during 2022, in a few months has shown huge potential from an economic, cultural and social point of view. , as shown Many discussions about jobs – Also in the creative and highly specialized field – which is He could fall into a crisis from its widespread implementation.

Thus, what unites the two phenomena is not a technological aspect but the so-called “hype cycle” (or “hype”), methodology Developed by technology analysis and consulting firm Gartner to graphically represent the development and adoption of new technologies. The cycle consists of five stages: the initial trigger, that is, the presentation of the technological innovation, followed by the media attention, thanks to which the pinnacle of exaggerated expectations, the moment of fame and optimism is reached. From here one plunges into the pit of disappointment, when experimentation does not lead to the desired results, and easy optimism leaves room for disappointment and abandonment for many users and entrepreneurs. We go way back through the rise of the Enlightenment, when consumers and investors realized that technology could have applications that could make it economically sustainable, to the productivity plateau, when awareness spread and led to mass adoption.

(wikipedia)

Not all technological innovations are able to complete the cycle. In many cases, there is no real, cost-effective application of the technology, or it is not achievable on a large scale. This is the state of the metaverse, the combination of virtual reality and social networks proposed by Meta and other companies starting in 2021, which according to some experts will not yet be sustainable with today’s technological level.

It is also due to the hype cycle that digital companies are constantly looking for the “next big thing”, the next innovation capable of changing the sector (or the world at large) forever. The perfect example of this phenomenon is Steve Jobs’ famous presentation of the iPhone in 2007, an event so significant that it inspired the expression “iPhone moment”, which refers to a single event capable of changing the fate of the industry. In recent weeks, several publications have called ChatGPT’s success an “AI iPhone moment”: even market analyst Rowan Curran has He said to the site VentureBeat That “the only thing that can compare to” OpenAI is precisely an Apple device. “What’s really amazing is that we have technology that’s useful today, that’s advancing quickly, and that we’re learning about in real time,” he explained.

However, AI systems seem to follow a different pattern than the traditional cycle. Even the spread of ChatGPT was special, far from the hype model suggested by Gartner: “Within a few days there were 1 million users,” Curran explained. “Even if a quarter of them are actually duplicates, there are still hundreds of thousands of human brains that suddenly find themselves playing with this technology,” which is something very different from usual. Its strange nature is also shown by the reactions it has elicited from many users and onlookers, who were so surprised by its functionality that they compared it to something magical.

The somewhat magical aura that surrounds the industry is also a lucky marketing accident that contributes to the promotion of ChatGPT and similar products. Jobs himself, while introducing the iPhone, defined the operation of the device in similar words: “It works like magic.” That artificial intelligence, according to New York MagazineCome Tell “As a story about inescapable technologies that are so transformative, so incomprehensible, and so unexpected that they preemptively return humanity to a pre-modern state of awe and mysticism.” A phenomenon already studied in 2020 by researchers Alexandre Campolo and Kate Crawford, who noted how the positive results obtained by technology Deep learning It generated “a cornucopia of optimistic discussion describing these systems as magical, appealing to mystical powers and supernatural powers”.

OpenAI itself, the company that created ChatGPT and DALL-E, was founded in 2015 as a non-profit organization with the grand mission of promoting and developing “friendly” AI so that humanity can benefit from it. At the time, co-founder Sam Altman said that the advent of AI “will likely bring about the end of the world but will allow great companies to be born,” showing that he believes in the uncanny and magical potential of this technology. In 2019, OpenAI changed its company name, become A for-profit company now worth about $29 billion.

However, not everyone is convinced of how these products are described. In recent days, Yann LeCun, Head of Artificial Intelligence at Meta, introduced the group that includes Facebook, Instagram and WhatsApp, Interview in which he defined ChatGPT as “not particularly innovative” and “nothing revolutionary, even if that is what the public experiences”. After the article was released, LeCun attempted to explain himself on Twitter, stating that he did not want to criticize the company’s work but rather to “correct the general perception of the public and media that ChatGPT is this unique and great innovation before all that remains.” It’s not really like that.”

To be clear: I am not criticizing OpenAI’s work, nor their claims.

I’m trying to correct the *perception* by the public and the media who see chatGPT as this new, incredibly innovative, unique technological breakthrough that’s far superior to anyone else’s.

it’s not like that.

– Yann LeCun (@yelecun) January 24, 2023

Commenting on the controversy, AI expert Alberto Romero said underlined How the fact that generative AI systems are “measurably useful” doesn’t make their hype less of an issue, just because of the comparison to Web3. Romero particularly criticized “AI influencers”, personalities, and investors who tend to exaggerate and embellish the truth, stating that “exaggeration that causes confusion is much more difficult than fighting the lies that lie under their weight.”