Artificial intelligence. Freefall. Dzhimsher Chelidze

business investment will decrease, and there will be questions about its feasibility.

      For example, on June 16, 2024, Forbes published an article: “Winter of artificial intelligence: is it worth waiting for a drop in investment in AI”.

      The original article is available by using a QR-code and a hyperlink.

      Winter of artificial intelligence: is it worth waiting for a drop in investment in AI

      It provides interesting analytics about the winter and summer cycles in the development of AI. Also included are the opinions of Marvin Minsky and Roger Schank, who back in 1984 at a meeting of the American Association for Artificial Intelligence (AAAI) described a mechanism consisting of several stages and resembling a chain reaction that will lead to a new winter in AI.

      Stage 1. The high expectations of business and the public from artificial intelligence methods do not justify themselves.

      Stage 2. Media outlets start publishing skeptical articles.

      Stage 3. Federal agencies and businesses reduce funding for scientific and product research.

      Stage 4: Scientists lose interest in AI, and the pace of technology development slows down.

      And the experts ' opinion came true. For a couple of years, the AI winter has been on the rise, and it only warmed up in the 2010s. Just like in “Game of Thrones”.

      Now we are at the next peak. It came in 2023 after the release of ChatGPT. Even in this book, for the reader’s understanding, I often give and will continue to give examples from the field of this LLM, although this is a special case of AI, but it is very clear.

      Further, the article provides an analysis of the Minsky and Schank cycle to the current situation.

      “Stage 1. Business and public expectations.

      It is obvious to everyone that the expectations of the revolution from AI in everyday life have not yet been fulfilled:

      – Google has not been able to fully transform its search. After a year of testing, the AI-supercharged Search Generative Experience technology receives mixed user reviews.

      – Voice assistants (“Alice”, “Marusya”, etc.) may have become a little better, but they can hardly be called full-fledged assistants that we trust to make any responsible decisions.

      – Support service chatbots continue to experience difficulties in understanding the user’s request and annoy them with inappropriate responses and general phrases.

      Stage 2. Media response.

      For the AI bubble query, the” old” Google search returns articles from reputable publications with pessimistic headlines:

      – The hype bubble around artificial intelligence is deflating. Difficult times are coming (The Washington Post).

      – From boom to boom, the AI bubble only moves in one direction (The Guardian).

      – Stock Market Crash: A prominent economist warns that the AI bubble is collapsing (Business Insider).

      My personal opinion: these articles are not far from the truth. The market situation is very similar to what it was before the dot-com crash in the 2000s. The market is clearly overheated, especially since 9 out of 10 AI projects fail. Now the business model and economic model of almost all AI solutions and projects is not viable.

      Stage 3. Financing.

      Despite the growing pessimism, we cannot yet say that funding for AI developments is declining. Major IT companies continue to invest billions of dollars in technology, and leading scientific conferences in the field of artificial intelligence receive a record number of applications for publication of articles.

      Thus, in the classification of Minsky and Schank, we are now between the second and third stages of the transition to the winter of artificial intelligence. Does this mean that “winter” is inevitable and AI will soon take a back seat again? Not really”.”

      The article concludes with a key argument – AI has penetrated too deeply into our lives to start a new AI winter:

      – facial recognition systems in phones and subways use neural networks to accurately identify the user.

      – Translators like Google Translate have grown significantly in quality, moving from classical linguistics methods to neural networks.

      – modern recommendation systems use neural networks to accurately model user preferences.

      Especially interesting is the opinion that the potential of weak AI is not exhausted, and despite all the problems of strong AI, it can be useful. And I fully agree with this thesis.

      The next step in the development of artificial intelligence is the creation of newer and lighter models that require less data for training. You just need to be patient and gradually learn the tool, forming competencies in order to use its full potential later.

      Chapter 5. AI Regulation

      The active development of artificial intelligence (AI) leads to the fact that society and states become concerned and think about how to protect it. This means that AI will be regulated. But let’s look at this issue in more detail, what is happening now and what to expect in the future.

      Why is AI development a concern?

      What factors are causing a lot of concern among states and regulators?

      – Opportunities

      The most important point that all of the following will rely —on is opportunities. AI shows great potential: making decisions, writing materials, generating illustrations, creating fake videos -you can list them endlessly. We don’t yet realize all that AI can do. But we still have a weak AI. What will general AI (AGI) or super-strong AI be capable of?

      – Operating mechanisms

      AI has a key feature —it can build relationships that humans don’t understand. And thanks to this, he is able to both make discoveries and frighten people. Even the creators of AI models do not know exactly how the neural network makes decisions, what logic it obeys. The lack of predictability makes it extremely difficult to eliminate and correct errors in the algorithms of neural networks, which becomes a huge barrier to the introduction of AI. For example, in medicine, AI will not soon be trusted to make diagnoses. Yes, they will make recommendations to the doctor, but the final decision will be left to the individual. The same applies to the management of nuclear power plants or any other equipment.

      The main thing that scientists worry about when modeling the future is whether a strong AI will consider us a relic of the past?

      – Ethical component

      There is no ethics, good or evil for artificial intelligence. There is also no concept of “common sense” for AI. It is guided by only one factor – the success of the task. If this is a boon for military purposes, then in ordinary life it will frighten people. Society is not ready to live in such a paradigm. Are we ready to accept the decision of an AI that says that you don’t need to treat a child or you need to destroy an entire city to prevent the spread of the disease?

      – Neural networks can’t evaluate data for reality and consistency

      Neural networks simply collect data and do not analyze facts or their connectedness. This means that AI can be manipulated. It depends entirely on the data that its creators teach it. Can people fully trust corporations or start-ups? And even if we trust people and are confident in the interests of the company, can we be sure that there was no crash or data was not “poisoned” by intruders? For example,