Shulou(Shulou.com)11/24 Report--

Zuckerberg was questioned again by members of Congress, but this time the entire AI open source community is on his side. He also talked about the significance of AI open source and his views on the future development of AI in the talk show.

Xiao Za has made a big deal again recently.

Two members of Congress, Richard Blumenthal (chairman of the Congressional Privacy, Technology and Law Committee) and Josh Hawley (a member of the Congressional Privacy, Technology and Law Committee), wrote a harshly worded inquiry letter to Xiaoza.

The two MPs wrote a letter severely questioning Xiaoza about the LLaMa leak and asked Xiaoza to answer several key questions before the 15th in response to the previous LLaMa model leaked in 4chan.

The most serious of these are these three:

Did Meta conduct a risk assessment of leaks and other situations before releasing LLaMa?

After the release of LLaMa, did Meta do anything to prevent the model from leaking, or reduce the harm caused by the model leak?

Does Meta use user data to train models or conduct research on AI techniques?

The whole letter is very strongly worded, full of crimes, Internet fraud and various negative situations that may be caused by the disclosure of LLaMa.

It is mentioned in the letter:

"We are concerned about the leakage of Meta's AI model, and we are very concerned about Meta AI (LLaMA) and its possibility of abuse and harm in spam, fraud, malware, invasion of privacy, harassment and other misconduct."

"choosing to distribute LLaMA in such an unrestricted and irresponsible manner raises important and complex questions: when and how to publicly publish complex AI models is appropriate."

It seems to imply that if something big happens in the future, you must be unable to handle it.

And the letter for OpenAI and its product ChatGPT, but created a "Weiguangzheng" positive image.

"OpenAI's ChatGPT will reject the request in accordance with its code of ethics. By contrast, LLaMA will provide the requested letters, as well as other answers relating to self-harm, crime and anti-Semitism. "

This Richard Blumenthal was the most important questioner and moderator when Sam Altman was questioned in Congress some time ago.

His attitude towards Altman was quite agreeable at that time.

Compared with Xiaoza's previous question session, he not only interrupted Xiaoza's speech, but also held up a tit-for-tat sign with a collection of Xiaoza's apologies, questioning him why he only admitted his mistakes but did not repent.

Why is there such a big difference in treatment when they are also technology leaders?

Obviously, his model is open source for use by the AI research community, promoting the development of security technologies such as AI alignment and transparency, and is being questioned by old men and women in Congress every day.

The open source of competitors and the legitimacy of training data have been ignored by regulators and treated as guests of honor.

Coupled with the trampling on OpenAI and Meta in the inquiry letter, we can see that the regulators really have a great opinion on Xiaoza and Meta.

Technological backwardness can be caught up through hard work, and it will be picked on by regulators wearing tinted glasses, and it will be really troublesome after that.

At this time, Xiaozha's mind should be 10, 000 LLaMA models running past.

Zuckerberg, the hero of AI open source, but when it comes to the release of LLaMa, Zach and Meta are really recognized by the industry and the public.

On Richard Blumenthal's own Twitter, the main author of the inquiry letter, he retweeted his own letter and gave a solemn account of his actions.

But the following reply was almost overwhelmingly questioned or even attacked by netizens.

Netizens responded that open source code is protected by the first Amendment. As long as open source Meta has no way to prevent it from being used by the public, so you ask these questions and I doubt your motives.

Who gave you the right to think AI was dangerous? Have people ever voted for you to stop AI technology from being open source and free? As a representative of the people, you have crossed the line!

Meta open source, does not involve commercial interests, then "Commerce Clause" can not control! It was leaked by someone outside Meta. What's the use of going to Meta?

Members, you really do not have the power to monitor the development of AI, neither now nor in the future. And how can it be completely controlled if everyone in the world has access to public web services?

Please use your legislative power to ensure that your job as a member will not be replaced.

I hope you will be chosen in the future.

LLaMa is just a footnote. Falcon in the United Arab Emirates is commercially open source directly, and more projects will be based on these open source models. So please don't stifle American innovation and key open source technologies in order to pursue impossible goals.

There is no need to blame FB, which will force companies that want to open source to shut down, making the technical transparency we pursue even more difficult to achieve. This is a very dangerous situation.

There is no point in letting Meta bear the consequences of AI abuse. Meta does not have the ability or responsibility to bear this, just as it is for the same reason that manufacturers of guns, hammers and baseball bats cannot be asked to bear the consequences of such abuse.

Among the 110 responses at the bottom of his tweet, there were few messages supporting him in writing the letter.

It can be seen that most of the netizens who left messages are explaining that the release of LLaMa and the open source of large language models are driving the progress of the industry and technology.

Because for the AI industry, open source is probably the most important way for the public to fully understand this technology and to ensure that AI technology securely serves mankind.

If one day Xiaoza or Meta were to be blamed for AI security, the logic behind it would be as absurd as Gotham's attempt to keep Gotham safe by blaming Batman.

Therefore, this accountability letter from members naturally stands on the opposite side of the public and is unpopular.

Meta's support on LLaMa comes not only from the public, but also from the open source community.

Some even compare the open source of LLaMa to the release of Linux.

Because of the open source operating system Linux, the threshold for web startups has been effectively lowered.

Data centers can run on commercial hardware and software, providing the necessary conditions for the birth of cloud technology.

It was the leak of LLaMa on 4chan that led to the "alpaca explosion" in March.

It makes it possible for AI's open source community model to compete with closed-source technology giants like Microsoft and Google.

If, like the original software industry, we divide the current AI industry into two camps: open source and closed source.

Meta, led by Xiaoza, is the leader of the open source camp.

Xiaoza's AI ambition just a few days ago, Xiaoza and Lex Fridman, a famous Silicon Valley technology talk show host, conducted a nearly three-hour interview.

In the interview, he fully shared his vision for AI technology and industry.

Maybe we can see why he is betting the future of Meta AI on open source and the open source community.

Future AI requirements determine that open source is the answer to the problem. In Xiaoza's view, future users will interact with many different AI, just like now you want to use different App to accomplish different tasks.

It's like making different emotional connections with different people in life.

This demand also exists in the business world, and every company in the world will want to have an AI.

Just like you have your own home page on Instagram, Facebook, WhatsApp or other platforms, you want to direct people to AI with which they can interact.

In this way, AI will recommend more of its own products to users than those of competitors.

Therefore, it is impossible for a single AI to solve all problems in the future, because it may be inconsistent with the interests of each particular enterprise.

This is why Xiaoza believes that there must be a clear need for different AI in the market and in life.

In order to meet the demand for diversified and customized AI in the future, open source is the best answer.

Clarifying why it is necessary to open the source, Xiaoza further shared:

Open source can generate more innovation within the ecosystem, make Meta products better, and provide valuable feedback on security for Meta, which is very important for Meta to build a good system.

Another reason is that open source software tends to be more secure, because it is easier to find vulnerabilities with more censors.

At the current AI scale, it is easier to align and understand through AI open source.

These are the benefits that AI's open source will bring to the AI industry and Meta itself.

AGI era faces new choices from the perspective that it has not yet reached the AGI stage, open source is a better choice than closed sources in a more open environment.

But in the era of approaching AGI or super artificial intelligence, this problem needs to be reconsidered and discussed.

Xiaoza believes that organizations and structures that transcend human intelligence already exist in the world.

For example, as an entity, the company has its own unique brand.

As a company, Meta can embody intelligence beyond the individual.

Similar to a company's network system, it combines the intelligence of multiple people, or the intelligence of multiple people and technologies, to form an entity that is much smarter than any other person in the world. and be able to use these systems to have an impact on society in a positive way.

No one knows when AGI will appear, but we can recognize from history that this more intelligent system is a beneficial phenomenon to promote human progress.

To some extent, individual AI systems make the interaction between individual human beings and other people more intelligent, so that collective intelligence is also more intelligent.

At the end of the sharing, Xiaoza didn't forget to laugh at himself.

"I have experienced too much failure and embarrassment in my life, and enduring failure and embarrassment has become my core competitiveness," he said.

Reference:

Https://venturebeat.com/ai/senators-send-letter-questioning-mark-zuckerberg-over-metas-llama-leak/

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.