In 2025 the most powerful AI ever brought us: What can we do with these charged skills (and what’s next)?

The latest AI major language model (LLM), like Claude 3.7 by Anthropic and Grok 3 of Xai, are published often list for doctoral students – at least according to certain benchmarks. This performance marks the next step towards what the former Google CEO Eric Schmidt imagine: A world in which every access to “large polymath” has a AI that can rely on huge knowledge strips to solve complex problems across disciplines.

Professor of the Wharton Business School, Ethan Mollick A useful thing Blog that these latest models were trained with significantly more computer performance than GPT-4 at the start two years ago. Grok 3 was trained up to ten times as much calculation. He added that this would make Grok 3 the first “Gen 3” KI model and emphasizes that “this new generation of AIS is smarter and the jump in the ability is striking.”

For example, Claude 3.7 shows emerging skills such as the expectation of user needs and the ability to take new angles into account in problem solving. After Anthropic it is the first Hybrid Argumentation model combines a traditional LLM for quick reactions with advanced argumentation functions to solve complex problems.

Mollick attributed these progress to two convergent trends: the rapid expansion of computing power for the training of LLMS and the increasing ability of AI to tackle a complex problem solving (often described as an argument or thinking). He came to the conclusion that these two trends are “overcharging AI skills”.

What can we do with this charged AI?

In a significant step Openai launched His “Deep Research” agent in early February. In his review too PlatformCasey Newton commented that Deep Research seemed “impressively competent”. Newton found that Deep Research and similar instruments could significantly accelerate research, analysis and other forms of knowledge work, although their reliability in complex areas is still an open question.

Based on a variant of the still unpublished O3 argumentation model, deep research can be expanded for a long time. This is done using the thinking of the chain (cot) (cot), whereby complex tasks are divided into several logical steps, as a human researcher could refine his approach. It can also search the web and make it possible to access more current information than in the training data of the model.

Timothy Lee wrote in Ai understand Through several test experts, it led to deep research and found that “his performance shows the impressive skills of the underlying O3 model”. In a test, it was asked about instructions after building a hydrogen electrolysis system. A mechanical engineer commented on the quality of the edition and estimated that it would need an experienced professional per week to create something as good as that as that 4,000 words report Openai generated in four minutes. ”

But wait, there are more …

Google Deepmind recently also recently released “Ai-Co-Scientist”, a multi-agent AI system based on its Gemini 2.0 LLM. Scientists are intended to help create new hypotheses and research plans. Imperial College London has already proven the value of this tool. According to Professor José R. Penadés, his The team spent years Decryped why certain super bugs contradict antibiotics. AI replicated their results in just 48 hours. While the AI dramatically accelerated hypothesizing, human scientists still had to confirm the results. Still Penadés said The new AI application “has the potential to end science.”

What would it mean to end science?

Last October, anthropic CEO Dario Amodei wrote in his “”Machines of loving grace“Blog that he was expecting” powerful AI ” – his term for what most refer to what most described as artificial general intelligence (AGI) – would” lead the next 50 to 100 years of biological (research) progress in 5 to 10 years. “Four months ago, the idea of compressing up to a century of scientific progress in a single decade was extremely optimistic in view of the recent progress in AI models, Including anthropic Claude 3.7, Openai Deep Research and Google Ai-Co scientist, what Amodei calls short-term “radical transformation” looks much more plausible.

While AI can quickly pursue the scientific discovery, biology is still bound and experimental validation, regulatory approval and clinical studies to real restrictions. The question is no longer whether AI will change science (as it is certainly certainly), but how quickly its full effect is realized.

In February 9th in February 9th Blog Post, the CEO of Openaai, Sam Altman, claimed that “systems that indicate AGI are in sight.” He described Agi as “a system that can increasingly tackle more complex problems in many areas”.

Altman believes that reaching this milestone could unlock an almost utopian future in which “economic growth looks amazing in front of us, and we can now imagine a world in which we heal all the diseases, have a lot more time to enjoy with our families and to fully exploit our creative potential.”

A dose humility

This progress of the AI is very important and ensure a completely different future in a short time. However, the meteoric rise of AI was not without a stumbling block. Consider the recent fall of the human AI pencil Ted talk. Almost a year later, the company collapsed and its Remains were sold From for a fraction of your once high rating.

For many reasons, ARI applications in the real world are often exposed to considerable obstacles, from lack of specialist knowledge to restrictions on the infrastructure. This was certainly the experience of Sensei AG, a startup that was supported by one of the richest investors in the world. The company wanted to apply AI by breeding improved harvesting types to agriculture and use robots for harvesting, but has met major hurdles. After For the Wall Street Journal, the startup has exposed many setbacks to technical challenges to unexpected logistical difficulties that highlight the gap between the potential of AI and its practical implementation.

What’s next?

When we look in the near future, science is about to discover a new golden age, with the AI becoming an increasingly capable partner in research. Deep learning algorithms that work together with human curiosity could uncover complex problems at the speed of recording when AI systems recognize large data trudes seven, invisible patterns for humans and interdisciplinary hypotheses.

Scientists are already using AI to compress research periods, predict the protein structures, scan literature and reduce the years of work to months or even days – to unlock the opportunities in terms of areas from climate science to medicine.

However, since the potential for radical transformation becomes clearer, but also the impending risks of disorders and instability. Altman himself admitted in his blog that “the balance of power between capital and work could easily get mixed up”, a subtle but considerable warning that the economic effects of AI could be destabilizing.

This concern has already been materialized, as shown in Hong Kong, how the city has recently Cut 10,000 The public service starts with AI investments at the same time. If such trends are continued and becoming more expansive, we could increase widespread recovery workers, increase social unrest and increase intensive pressure on institutions and governments worldwide.

Adaptation to a AI-operated world

The growing skills of AI in relation to scientific discovery, argument and decision -making mark a profound shift, which is both unusual promises and impressive challenges. While the path forwards can be characterized by economic disorders and institutional tribes, the story has shown that companies can adapt to technological revolutions, if not always simply or without consequence.

In order to successfully control this transformation, companies in governance, education and workforce must invest in order to ensure that the advantages of AI are fairly distributed. Even if the AI regulation is confronted with political resistance, scientists, political decision -makers and managing directors have to work together to build ethical framework conditions, enforce transparency standards and craft guidelines, reduce the risks and at the same time increase the transformative effects of the AI. If we face this challenge with foresight and responsibility, people and AI can master the biggest challenges of the world and initiate a new age with breakdowns that once seemed impossible.