Center towards AGI: How argument and deep research extend the AI ​​from statistical prediction to structured problem solving

AI has developed with one astonishing pace. What seemed like science fiction a few years ago is now an undeniable reality. My company started a AI excellence center in 2017. AI was certainly better used in predictive analysis and many algorithms for machine learning (ML) were used for speech recognition, for spam detection, to check magic sayings (and other applications) – but it was early. We then believed that we were only in the first inning of the AI ​​game.

The arrival of GPT-3 and in particular from GPT 3.5, which is set for the use of conversation and as the basis for the first chatt in November 2022, a dramatic turning point serves, which was always recalled as a “chatt-moment”.

Since then there has been an explosion of the AI ​​skills of hundreds of companies. Openai GPT-4 published in March 2023, which promised “Spark from Agi”(Artificial General Intelligence). At this point it was clear that we went far beyond the first inning. Now it feels like we’re in the last route of a completely different sport.

The flame of agi

The flame of Agi appears two years later.

In a recent episode of the hard fork PodcastDario Amodei, which has been working in the KI industry for a decade, formerly as a VP of Research at Openai and now as CEO of Anthropic -said that there is a chance of 70 to 80% that we are a “very large number of AI systems that are much intelligent as humans, almost everything before the end of the decade and my presumption 2026 or 2027.”

The evidence of this prediction becomes clearer. End of last summer, Openaai started O1 – The first “argumentation model”. They have published O3 since then, and other companies have introduced their own argumentation models, including Google and, as is well known, Deepseek. Reason use the chain (cot) that divide complex tasks into several logical steps at the term, just like a person can tackle a complicated task. Ki-Helized AI agents, including Openais Deep Research and Google’s AI CO scientists, have recently appeared, which have great changes to the execution of the research results.

In contrast to previous large -scale models (LLMS), which were mainly adapted from training data, they mainly made patterns, Argumentation models a fundamental shift of statistical prediction for structured problem solving. This enables AI, new problems go beyond his training and enable real argumentation instead of advanced pattern recognition.

I recently used Deep Research for a project and was reminded of the quote from Arthur C. Clarke: “Every sufficiently progressive technology cannot be distinguished from magic.” In five minutes, this AI produced what would have taken me 3 to 4 days. Was it perfect? Was it close? Yes very. These agents quickly become really magical and transformative and are among the first of many similarly powerful agents that will soon come onto the market.

The most common definition of AGI is a system that can perform almost all of them Cognitive task A person can do it. These early agents of the change suggest that Amodei and others who believe that we are close to this level of AI raffinness could be correct and that Agi will soon be here. This reality will conduct too much change and adapt people and processes in a short time.

But is it really agi?

There are different scenarios that could emerge from the short -term arrival of powerful AI. It is challenging and scary that we don’t really know how to do it. New York Times The columnist Ezra Klein addressed this in A Newer podcast: “We rush on AGI, without really understanding what that is or what that means.” For example, he claims that it would really mean little critical thinking or emergency planning in relation to the effects and, for example, for employment.

Of course, there is another perspective on this uncertain future and lack of planning, as illustrated by Gary Marcus, which will not lead to deep learning in general (and LLMS especially). Marcus output What corresponds to a setting of Klein’s position quotes remarkable defects in the current AI technology and indicates that it is also likely that we are a long way from AGI.

Marcus may be correct, but this could also simply be an academic dispute over semantics. As an alternative to the AGI Blogas a similar idea conveys without the inaccurate definition “sci-fi luggage and hype”. Name what you want, but AI will only be more powerful.

Play with fire: The possible AI Futures

In A 60 minutes interviewSundar Pichai, CEO of Alphabet, said he sees AI as “the most profound technology that works on humanity. Deeper than fire, electricity or anything we have done in the past. “This certainly fits the growing intensity of AI discussions. Fire like AI was a worldwide -changing discovery that progressed progress, but demanded control to prevent a disaster. The same sensitive balance applies today for AI.

A discovery immense and fire instruments changed civilization by enabling warmth, cooking, metallurgy and industry. But it also brought destruction when they were uncontrolled. Whether AI will make our greatest ally or our cancellation depends on how well we deal with the flames. To get this metaphor further, there are different scenarios that could soon emerge from an even more powerful AI:

1. The controlled flame (utopia): In this scenario, AI is used as a force for human prosperity. Productivity sprays, new materials are discovered, personalized medicine is available for everyone, goods and services are plentiful and inexpensive and individuals are freed from placking in order to pursue more meaningful work and activities. This is the scenario that is used by many accelerators in which AI makes progress without getting involved in too much chaos.
2. The unstable fire (challenge): Here AI brings undeniable advantages-revolutionized research, automation, new skills, products and problem solutions. However, these advantages are distributed unevenly – while some thrive, but others are exposed to a shift, expand economic distinctions and emphasize social systems. Mounting and security risks Mount. In this scenario, society struggles to reconcile promise and danger. It could be argued that this description of today’s reality is close.
3. The forest fire (dystopia): The third way is a disaster, the possibility that is most associated with so -called “doomers” and “probability of fate”. Whether by unintentional consequences, ruthless use or AI systems that go beyond human control, AI actions are not checked, and accidents occur. Trust in the truth eroded. In the worst case, AI is shirking out of control and threatened living, industries and entire institutions.

While each of these scenarios seems plausible, it is uncomfortable that we really don’t know which are most likely, especially since the timeline could be short. We can see early signs of each: AI-controlled automation increases productivity, misinformation that spread on the scale, erode trust and concerns about insignificant models that resist their guardrails. Each scenario would cause its own adjustments for individuals, companies, governments and society.

Our lack of clarity about the trajectory for AI effects indicates that a mix of all three futures is inevitable. The rise of AI will lead to a paradox that accepts prosperity and at the same time has unintentional consequences. Astonishing breakthroughs will take place, as well as accidents. Some new fields will appear with tempting opportunities and professional prospects, while other staled companies will have an impact on the economy.

We may not have all answers, but the future of the powerful AI and its effects on humanity are now being written. What we saw at the latest action summit in Paris Ai was the attitude to hope for the best thing that is not an intelligent strategy. Governments, companies and individuals have to shape the KI trajectory before forming us. The future of the AI ​​is not determined solely by the technology, but by the collective decisions that we make about the provision.

Gary Grossman is EVP of technology practice at Noble.