2025 has already brought us the most performant AI ever: What can we do with these supercharged capabilities (and what’s next)?

The latest AI Big Language Model (LLM) releases, such as Claude 3.7 from Anthropic, Grok 3 from XAI, are Performance frequently On a PhD – at least according to certain benchmarks. This achievement marks the next step for former Google CEO Eric Schmidt Imagine: A world where everyone can use “great polymorphism”, one’s AI can leverage a huge body of knowledge to solve complex interdisciplinary problems.

Wharton professor Ethan Mollick A useful thing The blog says these latest models were trained using greater computing power than the GPT-4 when they were released two years ago, and the Grok 3 is 10 times more trained than computing. He added that this will make the Grok 3 the first “Gen 3” AI model, emphasizing that “the new generation of AIS is smarter, and the jump in features is shocking.”

For example, Claude 3.7 shows emerging features such as anticipating user needs and the ability to consider new perspectives for problem solving. According to the personification, this is the first Hybrid Inference model, combining traditional LLM with fast-responsive traditional LLM with advanced reasoning capabilities to solve complex problems.

Mollick attributes these advances to two fusion trends: the rapid expansion of computing power trained by LLM and the improvement of AI’s ability to solve complex problems (often described as reasoning or thinking). He concluded that these two trends are “enhanced AI capabilities.”

How should we deal with this supercharged AI?

In an important step, Openai emission Its “deep study” AI agent was in early February. In his comments Platformer GamesCasey Newton commented that in-depth research seems to be “impressive”. Newton notes that in-depth research and similar tools can significantly accelerate research, analytical and other forms of knowledge work, although their reliability in complex areas remains an open question.

Based on variants of the still unissued O3 inference model, in-depth research on inferences that can be extended over a long period of time. It does this using chain of thought (COT) reasoning, breaking complex tasks into multiple logical steps, just as human researchers might perfect their approach. It can also search the network, allowing it to access more latest information than in model training data.

Timothy Lee Understand AI A test conducted in-depth research on several experts and noted that “its performance demonstrates the impressive functionality of the underlying O3 model.” One test asked how to build a hydrogen electrolytic plant. A mechanical engineer commented on the quality of output, “Estimated every week requires experienced professionals to create with 4,000-word report Openai is generated in four minutes. ”

But wait, there are more…

Google Deepmind has also been recently issued “AI Co-scientist”, a multi-agent AI system based on Gemini 2.0 LLM. It aims to help scientists create novel hypotheses and research programs. Imperial College London has proven the value of the tool. According to José R. Professor Penadés’ statement The team has spent many years Reasons for dissolving certain superbacterial antibiotics. AI copied their findings in just 48 hours. Although AI significantly accelerates the generation of hypotheses, human scientists are still needed to confirm these findings. However, Pengades explain New AI applications have the potential to enhance science.

What does this mean for joining science?

Last October, Human CEO Dario Amodei was in his “The machine of grace“He expects that the blog “strong AI” (what he calls artificial general intelligence (AGI)) will lead to “the next 50 to 100 years of biology [research] Progress within 5 to 10 years. “The idea of ​​taking a century of scientific advancement to a decade of scientific advancement four months ago seemed very optimistic. With the latest advances in AI models, including the anthropomorphic Claude 3.7, Openai Deep Research, and Google AI Co-Scientist, Amodei’s known as the recent “radical transformation” is starting to look more reasonable.

But while AI may make scientific discoveries quickly, at least biology remains constrained by reality-experimental verification, regulatory approval, and clinical trials. The question is no longer whether AI will change science (certainly), but that it can achieve all its impact.

On February 9 blog Post, Openai CEO Sam Altman claims that “systems that start pointing toward AGI are unveiling.” He describes AGI as “a system that can solve increasingly complex problems at the human level in many areas.”

Altman believes that achieving this milestone can unlock a near-utopian future, “The economic growth ahead of us looks amazing, and we can now imagine a world where we can cure all diseases, have more time to enjoy with our family, and can fully realize our creativity.”

humble

These advances in AI are very important and predict a very different future in a short period of time. However, the rapid rise of artificial intelligence is not without accidents. Consider the recent fall of Humane AI PIN – a device that was hyped as a smartphone replacement after BuzzWorth Ted Talk. Just a year later, the company collapsed, its Residues are for sale With a small part of its once 50-degree valuation.

Real-world AI applications often face significant obstacles due to lack of relevant expertise to infrastructure constraints. Of course, this is the experience of Sensei AG, a startup backed by one of the world’s wealthiest investors. The company set out to apply AI to agriculture by breeding improved crop varieties and using robots to harvest them, but encountered major obstacles. according to For the Wall Street Journal, the startup faces many setbacks, from technical challenges to unexpected logistical difficulties, highlighting the gap between AI’s potential and actual implementation.

What’s next?

As we come, science is at the forefront of a new golden age of discovery, and AI has become an increasingly capable partner in research. As AI systems sift through large amounts of data, invisible spot patterns of humans and propose interdisciplinary assumptions, deep learning algorithms that work with human curiosity may reveal complex problems at record speed.

Scientists have used AI to compress research timelines—predicting protein structures, scanning literature, and reducing years of work to months or even days), thus unlocking opportunities in the fields from climate science to medicine.

However, as the potential for radical transformation becomes clearer, so are the risks of destruction and instability. Ultraman himself acknowledged in his blog that “the balance of power between capital and labor is easily messed up”, a subtle but significant warning that the economic impact of AI could destabilize.

As Hong Kong proves, this concern has been achieved Cut 10,000 Civil servants work while increasing investment in AI. If this trend continues and becomes more extensive, we can see widespread labor unrest, exacerbating social unrest and putting huge pressure on institutions and governments around the world.

Adapting to the AI-driven world

AI’s growing capabilities in scientific discoveries, reasoning and decision-making mark a profound transformation that not only raises extraordinary hopes but also brings great challenges. While the path forward may be characterized by economic disruption and institutional torture, history shows that society can adapt to technological revolutions, although not always easy or without consequences.

To successfully drive this transition, society must invest in governance, education and workforce adaptation to ensure that AI benefits are distributed equitably. Even if AI regulations face political resistance, scientists, policy makers and business leaders must work together to establish an ethical framework, implement transparency standards and formulated policies to mitigate risks while amplifying the transformative impact of AI. If we meet this challenge with vision and responsibility, then people and artificial intelligence can meet the biggest challenges in the world and usher in a new era with breakthroughs that were once impossible.