Gregory Gromov

The idea of the intellectual immortality of generations is knocking at the door of the next stage of the AI revolution. To give a first impression of what we are talking about, let me recall a counter-intuitive statistic on road safety by age groups: the safest drivers, by far, are seventy-year-olds. In other words, it takes people roughly half a century to truly learn how to drive safely. And then, all too soon, they pass away — with no opportunity to share that hard-won skill … not even with their own grandchildren, who must begin learning from scratch and often pay the ultimate price for their mistakes. This is not a unique situation. In fact, it is emblematic of a deeper human condition. All verbalized knowledge amounts to only a thin, transparent surface layer floating on the vast ocean of human knowledge & skills. That ocean is formed by a lifetime of accumulated experience, most of which remains inaccessible to others … felt only by the individual, stored in tactile, muscular, motor, and emotional memory rather than in any kind of sharable words. As for knowledge formalized in computers and other automated systems — that is merely a thin film on the surface of this immense ocean. Keeping all of the above in mind, I coined the term AUTOFORMALIZATION and described its meaning in an article: “Autoformalisation - Knowledge & skill acquisition” netvalley.com/cgi-bin/librar… From my more recent experience, I see that today’s most popular AI products — LLM services such as GPT — hold the potential to become effective tools for helping people extract their intuitive knowledge: first by verbalizing it, and then by autoformalizing it. Some people have asked me why I coined the term AUTOFORMALIZATION at a time when there were no tools capable of scaling the process it describes. My usual answer is a quotation from Heinrich Neuhaus book: “The Art of Piano playing” ‘To give a thing a name is the beginning of understanding it’

Not surprisingly, Iran’s Islamist revolution was literally imported from France and triumphed with the active support of a Democratic Party US president. Not to mention there were Pol Pot and many other monsters whose leftist ideologies were shaped in France.

As long as OpenAI’s leadership refuses to recognize the simple fact that chat features formed the conceptual foundation of the GPT moment, the company’s pace of innovation will continue to slow. Again, the strongest driver of the ChatGPT phase of the AI revolution was human–machine dialogue—not the search for solutions to Erdős problems or the other mathematical gymnastics Greg favors in the naive hope that the math key will open the doors to scientific discovery. Continuing to improve ChatGPT’s chat features will help scientists deepen their understanding and make new discoveries in the fields they care about. AI can be an effective tool in scientists’ efforts to explore new scientific horizons. Any attempt to turn the model itself into a scientist, legal expert, physician, or other professional is counterproductive at best.

So this began in the UK long before the Democrats’ recent landslide victory in the US midterms —and even before Biden’s disputed victory in 2020. Now the US is heading decisively down the same road as UK to nowhere. As Abdul El-Sayed, now Michigan’s senator-elect, said in 2018: “We as a country, I believe, owe it to the world to open our doors and we’re better because of it.”

The early foundation of the binary logic for kids:

France’s nuclear programme, whose foundations were laid under Charles de Gaulle and which expanded dramatically under the 1974 Messmer Plan, strengthened the country’s energy security and helped make it a major electricity exporter to neighbouring countries, including Germany. Nuclear power is among the lowest-carbon and safest sources of electricity.

While Trump tried to acquire the Canada, UK de facto acquired New Jersey:

AGI or No AGI—Is That Really the Question? Does solving an Erdős problem make AI more useful to a doctor, a biologist, or a writer? Perhaps—but the connection must be demonstrated. Before asking who is winning the AI race, we need to define what counts as progress. This is where the vocabulary of artificial general intelligence becomes less useful. If “closer to AGI” does not specify which capabilities have improved, under what conditions, and for whose purposes, it contributes little to evaluating a system. A doctor may need help identifying missing information or distinguishing evidence from speculation. A biologist may need assistance formulating a testable hypothesis or interpreting an unexpected result. A writer may want an interlocutor who can develop an idea while preserving an individual voice. Success in these activities cannot simply be inferred from success in proving a theorem. As practical uses multiply, the declining explanatory value of such undefined labels becomes predictable. Users need distinctions between the models effectiveness measured by domains that help them to choose the more effective AI tools for their needs from those available on the market The same objection applies to a collection of benchmarks presented as a universal measure of AGI wise progress. “Progress” in which field and to what extent? Domain-specific evaluations already exist. The proposal is to make them central to how progress is explained, compared, and funded. Each domain, and progressively each subdomain, needs tests that reflect its actual practices. Medical dialogue with patients, biological research assistance, creative writing, and the mathematical formalization of problems can require strikingly different evidence of success. Even a domain label can be too broad. Explaining a medical paper, organizing a patient history, and supporting diagnostic reasoning are different tasks. Helping a biologist design an experiment differs from summarizing published findings. An evaluation should make these distinctions visible rather than conceal them in a single score. The criteria must also reflect the consequences of failure. Accuracy, reliability, time saved, and cost matter in different proportions across tasks . Creative assistance requires judgments about originality and fit with the author’s intention. Research assistance requires attention to evidence and even greater attention to uncertainty. Meaningful comparison begins by stating these requirements. These distinctions should guide the allocation of scarce computing resources. Companies and research institutions should explain which capabilities they intend to improve and why those improvements deserve priority under their scientific, social, or economic objectives. Even the broadest comparisons require clearly stated criteria. If we mean web influence, we can examine audience reach, frequency of use, and engagement. If we mean commercial performance, we can compare revenue from a particular class of tasks over a defined period. These measures answer different questions and may help identify which frontier AI companies lead in the fields being compared. Neither establishes universal superiority under an abstract notion of AGI. The same discipline should govern technical evaluation. A mathematical achievement deserves recognition on mathematical grounds. Whether the methods behind it also improve assistance to doctors, biologists, engineers, or writers is a further and much more complex question, requiring evidence from tests within those domains. The underlying logic of an algorithm may transfer between domains, but the extent of that transfer must be measured rather than assumed. The practical task is therefore to establish what a substantial group of users within a chosen domain can accomplish, according to criteria they consider reasonable, how reliably, and at what cost—and to use that evidence to guide improvements in algorithms and the allocation of computing resources. Domain-specific benchmarks make this difficult task concrete. Broader comparisons can sometimes be useful, but only when their criteria are explicit. Before asking who is winning the AI race, we should specify which achievement we are measuring, for whom it matters, and why.

“Reality is usually less cinematic than the theories.” Grok remains in a stable position among 3 leaders of the AI race by the criteria of top benchmarks including the sense of humor

“Macron is constitutionally barred from seeking a third consecutive term, while far-right candidate Marine Le Pen is the favorite to win, the AP reported.”

A normal life of the normal people as a political platform of Trump’s new Chief of Staff is a reason the leftists are in panic:

Sources: