Why should we create AI tools, not act. | By vms | January 2025

Good afternoon! I am optimistic and I am here today to say that we can create an inspiring and transformer future by focusing on Tool Aiprovided you avoid building General artificial intelligence (AG). Acted, I think, is not only useless but also undesirable and avoidable.

Imagine a future where IA tools will help us to cure diseases, solve global challenges and improve our quality of life. Now ask yourself: Do we really need to do to get there? During the seminar, I asked the public two simple questions:

1. Do you want AI tools that help us cure diseases and solve problems?
(A sea of ​​hands got up.)

2. Would you prefer an AI that replaces us, makes us economically obsolete and potentially lead to our fall?
(Not a single hand has risen.)

This striking difference highlights a critical point: the AI ​​of which most of us dream of does not require act. Tool Ai is enough for almost all the transformative applications that we are considering.

Some maintain that we cannot benefit from the benefits of Tool AI without also developing AG. This idea is false. AGE, which refers to an intelligent AI than humans, capable of surpassing us in virtually all tasks, has significant risks. Imagine a system capable of building smarter robots, creating factories and reproducing endlessly. Such a system could evolve towards a new digital species – An area on which we could completely lose control.

It’s not just science fiction. Let’s make an analogy: imagine someone saying that we must accept Biotechnology totally unregulated With human cloning, eugenics and research on function gains – otherwise, we cannot have biotechnology at all. You would call Bs On this. Likewise, if someone tells you that we need Act not regulated To benefit from the advantages of AI, you must be skeptical.

The solution lies in Safety standards. As well as legally compulsory standards guarantee the safety of drugs, regulations may ensure that AI remains a tool and not a threat.

You might think that act is only a few decades and that it remains a distant concern. This is what most experts believed, until recently. In 2024, AI’s abilities exploded. Robots are now folding linen, generative AI systems provide transformative results and tools such as Chatgpt-4 Demonstrate an amazing mastery of language and knowledge. The forecasts concerning the deadlines of AC have radically changed.

For example:

• Dario AmodeiCEO of Anthropic, suggests that IGI could simply be in a few years.
• Sam AltmanCEO of Openai, said: “This is the first time that I have the impression that we really know what to do. From there, Building acted is only to do the work.

This collapse of the calendar is alarming, especially given our inability to control AG. Alan TuringIn 1951, warned against the control of the machines. He did not consider AI as a simple tool among others, but as a potential new species. A more intelligent species dominates naturally, just as humans dominate less intelligent species such as tigers.

Let me approach another myth: We need it to release all the potential of AI. It’s just not true. Here are some examples of what Tool Ai can achieve without act:

• Prevent accidents: Tool AI can save up to one million lives each year on the roads by preventing accidents.
• Advance health care: From prostate cancer From diagnosis to the folding of proteins for the development of drugs, Tool AI revolutionizes medicine.
• Education and sustainability: The tool that AI can democratize educationtransform industries and help us respond United Nations sustainable development objectives faster.

The main thing? The AI ​​tool is sufficient. Act is not necessary.

You can ask, If the act is inevitable, should we not focus on its control? Unfortunately, controlling AC is a problem that we have not resolved. And no, you don’t have to believe me. Many renowned experts share this concern.

• Alan Turing, Pioneer of AI Predicted that machines would take control once they go beyond human intelligence.
• Sam Altman and others warn that the act could lead to human extinction.

The problem is not whether act is aware or evil. It’s if it’s competent Enough to pursue objectives that conflict with ours. A missile in search of heat has no deep targets, but it behaves as if it was – and it is fatal. Likewise, AG does not need to “want” to hurt us; It is enough for him to act in a way that does not correspond to our values.

While most of us do not want act, a small minority defends it. For example:

• Professor Richard Sutton Consider AGA as the next step in evolution.
• BEFF JezosCreator of the E/ACC movement, openly supports the replacement of humanity with machines.

I call this ideology Digital eugenics – an alarming notion that machines are more “worthy” than humans. But I remain firmly on Human team. I want my one -year -old son to inherit a significant future, not dictated by the digital lords.

Finally, demystify the myth of inevitability. Humanity has already managed to ban harmful technologies – Human cloning And biological weaponsFor example. AG is no different.

The first step is to stigmatize Technology. If someone supports act, remind him that he may lose any agency for the benefit of machines. The second step is to implement Safety standards. Just as dangerous drugs are not marketed, Act should not be developed without guaranteeing its safety.

And no, it is not a question of competing with China. The greatest threat to security is not another country – it is an uncontrollable act. The Act race is not an arms race; It is a suicide race. Like the classic film War games Put it, “The only winning gesture is not to play.”

Why should we create IA tools, not acted

The debate on artificial intelligence (AI) has reached a pivotal moment. While the world is witness to rapid technological progress, the debate between IA-Outil and general artificial intelligence (AG) is becoming more and more urgent. AG, a system capable of performing any intellectual task that a human can achieve, has significant risks if it is not managed properly. On the other hand, Tool AI offers an inspiring vision of global prosperity without compromising control.

“The only winning gesture is not to play.” This deep declaration underlines why countries like the United States and China should avoid the perilous race towards AG. The inherent danger of AGE lies in its uncontrollable nature, which could deprive governments and societies of their autonomy. As a speaker pointed out: “No government, including the Chinese government, is encouraged to build an AGE which could remove all its power if it loses control. »»

Openai’s “Narrow path” plan (*control) offers a promising solution. This plan recommends treating AI as any other powerful technological industry, with binding safety standards. The United States and China must not only unilaterally apply these standards, but also collaborate on a moratorium treaty on AGS to guarantee that AGS are not developed in unleanished or thugs environments like North Korea.

Such security standards can lead to unprecedented innovation in Tool AI. As an expert said, “AI itself can help us create this AI tool”. By taking advantage of AI to create reliable and safe tools, we can make significant advances. The process would involve generating AI tools, writing code and producing verifiable evidence that these tools meet human specifications. “It is much more difficult to find proof than to check it after having found it”, which makes this approach both practical and safe.

The myth of Icarus is a timeless warning against pride. The speaker underlined: “AI gives us humans, amazing intellectual wings with which we can do things beyond the craziest dreams of our ancestors, provided that we do not waste it all in Obsessively trying to fly to the sun and build an act. The AI ​​tool, when developed in a responsible manner, can lead to a future of global prosperity without the existential risks posed by AG.

Even if the physical possibility of AGA is almost certain, its development depends entirely on societal decisions. “The question of whether we are going to really build it will simply depend on our decision, as a company, to want it or not.” The speaker compared the situation to a historical parallel: the creation by Enrico Fermi of the first autonomous nuclear reaction. Even if the technology itself was not immediately dangerous, it paved the way for the nuclear bomb in three years.

The modern equivalent, they say, occurred last year with the advent of large linguistic models. This “Enrico Fermi moment” means that humanity is closer than ever to Act. The question is not whether it is possible, but if we have the wisdom to delay its development until we can guarantee its security.

A notable part of the seminar underlined the central role of Elon Musk in the promotion of IA security. The long -term vision of Musk and his concerns concerning the risks of AI have led him to significantly support the IA security initiatives. As the speaker told: “Elon was already quite panicked at the time at the idea that we did something imprudent with AI. »The donation of $ 10 million Musk to the future of Life Institute helped generalize research on AI security, transforming it from a niche concern into a field of critical study.

The seminar ended with a powerful call for action: humanity must win the race for wisdom. “The key to a great future for high technology is to win the race between the growing power of technology and the wisdom with which we are the Orient towards good uses. By giving priority to security, collaboration and innovation in Tool AI, we can avoid the AGE traps and exploit technology for the common good.

In summary, the future of AI does not reside in the pursuit of the elusive dream of AG, but in the creation of reliable, safe and innovative AI tools that benefit humanity. As the speaker said, “It is now time to have this conversation”. The choices that we will make today shape the trajectory of AI and its impact on our world for future generations.

Before leaving: