Article reprint source: AI DreamWorks
Original source: New Wisdom
Image source: Generated by Unbounded AI
On October 9, Hinton was interviewed by "60 Minutes" and once again expressed his concerns that artificial intelligence might take over humans.
But as early as May this year, the "godfather of artificial intelligence" Geoffrey Hinton announced his resignation from Google in order to be able to talk freely about the risks of artificial intelligence.
Since then, Hinton has expressed his concerns about the risks of artificial intelligence in various public occasions.
In an interview, Hinton even said: "I deeply regret my life's work."
In this year when artificial intelligence is booming, it is difficult for people to stop charging forward.
Although at the end of March this year, thousands of technology leaders signed a joint letter calling for a moratorium on the development of artificial intelligence more powerful than GPT-4.
But at the YC alumni sharing session on September 24, Altman revealed news about OpenAI's next-generation large models GPT-5 and GPT-6.
At the same time, Meta, as a leader in the open source community, is also constantly developing new large models. In July, it released Llama 2, which has improved performance in all aspects and is close to ChatGPT.
As the performance of large models continues to improve, the shadow of security risks and threats always hangs over humanity like a sword of Damocles.
Some people think that it is too early to worry about the risks of artificial intelligence. For example, Yann LeCun, chief scientist of Meta and a disciple of Hinton, who is also a Turing Award winner, has repeatedly stated that people exaggerate the risks of artificial intelligence.
But some people insist that "no matter how serious the risks related to survival are, they cannot be taken too seriously."
Hinton is one of them.
Hinton: The possibility of AI taking over humans
In a conversation with a reporter from Wired, Hinton expressed his concerns about the risks of artificial intelligence from the following aspects.
The first is the growing intelligence of artificial intelligence.
Hinton believes that we are entering an era where artificial intelligence is becoming smarter than humans and is able to make decisions based on its own experience and understanding.
Although Turing Award winner Bengio and others used 88 pages to prove that artificial intelligence does not currently have consciousness, both Bengio and Hinton believe that artificial intelligence may develop self-awareness in the future.
In the future, artificial intelligence will replace humans and become the most intelligent being on earth.
What determines the intelligence of a model is its ability to learn. Hinton explains the learning process of artificial intelligence.
Artificial intelligence is able to learn because humans design the algorithms, but the neural networks that the models produce are complex and not fully understood.
Hinton likened this process to the principles of design evolution. He acknowledged that as artificial intelligence handles more complex tasks, its inner workings will become increasingly mysterious and unknowable to humans.
In addition, Hinton also expressed concerns that artificial intelligence systems will write their own code to modify themselves, which may cause humans to lose control of artificial intelligence.
In addition to the risks of AI getting out of control, Hinton discussed the risks of AI manipulating humans, pointing to potential consequences in areas such as autonomous battlefield robots, fake news, and employment bias.
But worse than all that, Hinton highlighted the uncertainty surrounding AI and the challenges of navigating uncharted territory.
“We are entering a very uncertain period, and we are dealing with things that have never been done before. Usually when you deal with something completely novel for the first time, you are bound to make mistakes. But we cannot make mistakes on these things.”
But Hinton also affirmed the contribution made by artificial intelligence.
With the various practical applications of artificial intelligence in the real world, it has greatly improved human efficiency in office, scientific discovery, medical care and other aspects.
Hinton particularly emphasized the huge advantages of artificial intelligence in the field of healthcare, especially medical image recognition and drug design. Alpha Fold launched by Google DeepMind is an excellent example.
Hinton expressed optimism about the positive impact of AI in this field.
Regarding the future of artificial intelligence, although Hinton did not predict that artificial intelligence will definitely take over humans, he admitted that this will be our possible future.
Therefore, Hinton has been urging humans to take necessary protective measures for their own safety in the face of the potential for the development of artificial intelligence.
Other concerns about the risks of AI
Previously, New York University professor Marcus expressed the same concern.
“Is it more important that humans are dying out, or that we can code faster and talk to robots?”
Marcus believes that although AGI is still too far away, LLM is already very dangerous. The most dangerous thing is that AI is now facilitating various criminal activities.
“In this era, we have to ask ourselves every day what kind of AI we want. Sometimes it’s better to go slower, because every small step we take will have a profound impact on the future.”
Sam Altman, the father of ChatGPT and CEO of OpenAI, has also publicly stated that the risks of super-intelligent AI may far exceed our expectations.
In a podcast with technology researcher Lex Fridman, Altman sounded the alarm about the potential dangers of advanced artificial intelligence.
He said what he is most worried about is the problem of false information caused by AI, the economic impact, and other problems that have not yet appeared. These harms may far exceed the level that humans can cope with.
He raised the possibility that large language models could influence, or even dominate, the experiences and interactions of social media users.
“For example, how do we know that big language models aren’t directing the flow of thought on Twitter?”
“I think we have a responsibility as OpenAI to be responsible for the tools that we put out into the world,” Altman said.
“AI tools will bring huge benefits to humanity, but, you know, tools have both benefits and disadvantages,” he added. “We will minimize the disadvantages and maximize the benefits.”
While the GPT-4 model is amazing, it’s also important to recognize the limitations of AI and avoid anthropomorphizing it.
Similarly, Bill Gates has been worried about the issue of "AI destroying humanity" for many years.
For years, Gates has said that artificial intelligence technology has become too powerful and could even start wars or be used to create weapons.
He claimed that not only could real-life "007 villains" use it to manipulate world powers, but artificial intelligence itself could also get out of control and pose a huge threat to humanity.
In his view, when superintelligence is developed, its thinking speed will far surpass that of humans.
It can do everything the human brain can do, and the most terrifying thing is that its memory size and operating speed are not limited at all.
These entities, called “super” AIs, may set their own goals and it’s unknown what they might do.
And the public is beginning to make their voices heard.
Faced with the great model "arms race" between large companies, Meta was met with citizen protests for opening up Llama 2's model weights to the public.
The public is concerned that Meta's strategy of releasing model weights would allow anyone to modify the model in their own interests.
At the same time, because the weights are open, it means that others cannot recall these models. This provides almost unlimited growth soil for using large models to commit cybercrime.
Now, for all mankind, it seems that "to be, or not to be" has become an imminent problem that must be faced.
References:
https://www.arabtimesonline.com/news/godfather-of-ai-fears-technology-could-one-day-take-over-humanity/