One Week AI Highlights: Suspend Super AI R&D? The multi-party game set off by the "joint letter"
After the AI explosion of "thousands of trees and pear blossoms", the game of public opinion triggered by the "joint letter" became the last hot topic in March.
Suspend super AI research and development for at least 6 months?
A "joint letter" became the protagonist of this week
On March 29 (Wednesday), a "joint open letter" written by the non-profit organization "Future of Life Institute" attracted global attention. In this joint letter, more than 1,000 bigwigs appealed that we should immediately Stop training AI systems stronger than GPT-4 for at least 6 months. At the beginning, there were many heavyweights in the signatures: Turing Award winner Yoshua Bengio, Apple co-founder Steve Wozniak, "A Brief History of Humanity" author Yuval Noah Harari, Musk and so on.
However, this "thousand-person joint letter" has only been published for less than a day, and different voices appeared. The Turing Award giant LeCun, who was originally reported as the "joint letter" by the platform, refuted the rumors on Twitter: No, I didn't sign it, and I disagree with the premise of the letter . They even "pierced the skylight and said the truth": the so-called "suspended R&D" is nothing more than "secret R&D", which is exactly the opposite of what some signatories hoped.
Today, some people have discovered that the signatures of some well-known people in the "joint letter" are fake, and some background of the source of the "joint open letter" "Future of Life Institute" has also been picked up. For example, Musk is the organization's support to which he donated $10 million in 2015.
Some people are questioning this "joint letter", while others are defending it. For the future of "Super AI R&D", only time will know the answer.
Midjourney suspends free trial
The influx of new users is the main reason
On March 28 (Tuesday), Midjourney CEO David Holz announced that he would stop the free trial service for new users. In an email to the media later, he explained the reason for the service stop: "It may be because an operation tutorial video is on the Internet in China. The virality of , the large number of people creating one-time accounts in order to get free pictures, this thing happened at the same time as the GPU shortage. The combination of these two things led to a degraded experience for paying users.”
At the beginning, many media people interpreted that the cessation of use was related to the use of Midjourney to generate "Trump was arrested" and other viral images that could be faked. However, Holz clarified that this is a "misunderstanding". After all, Midjourney's free trial is not the latest v5 version of Midjourney, which has not yet achieved the effect of confusing the public.
Midjourney was obviously a little caught off guard by the threat of AI images touching sensitive topics. They are currently working hard to improve their review policies. Many people have discovered that as the generated images spread wildly, Midjourney began to patch the list of taboo words.
Microsoft adds ads to Bing Chat
Official relaxation of the limit on the number of conversations
Bing Chat has had some good news and some bad news this week. First of all, Bing Chat has increased the current limit on the number of conversations per round to 20, with a maximum of 200 per day. The 50 limit is much better. At the same time, Microsoft is also testing the support of the "chat history" function in a small area to allow users to access previous chat records, which is a good optimization.
However, Microsoft's "small abacus" has also started fighting. The vice president of Microsoft revealed on March 29 (Wednesday) that "it is exploring the placement of advertisements in the chat experience", and shortly thereafter, some users reported that "the action is really fast" and they have seen advertisements.
Google Bard was exposed to ChatGPT data training
Or will join forces with DeepMind
According to The Information, former Google AI engineer Jacob Devlin believes that the Bard team is using information from ShareGPT, and the latter's users can post their communications with ChatGPT on the platform. He had tried to warn Google not to use ChatGPT data because it would violate OpenAI's terms of service and make Bard answers too similar to ChatGPT.
At present, the departing engineer has joined OpenAI, and a source revealed that Google stopped using data from ChatGPT after receiving a warning from Jacob Devlin. But a Google spokesperson stated that "Bard was not trained on any data from ShareGPT or ChatGPT."
Foreign media got news that the engineers of the Google Brain team are cooperating with DeepMind to develop a project "Gemini" that competes with OpenAI.
Japanese lawmaker ChatGPT questions the Prime Minister
And said that AI answers more sincerely
On March 29 (Wednesday), Japanese Prime Minister Fumio Kishida answered a question generated by Member Kazuma Nakatani using ChatGPT at a meeting of the House of Representatives.
Kazuma Nakatani asked ChatGPT about the issue of the "Proposed Amendment to the New Influenza Special Measures Act" in advance, "If you were a member of the House of Commons, what questions would you ask the Prime Minister?" ChatGPT asked: "The local government and healthcare professionals Have your comments been responded to? Could you please let me know how the relevant people have reacted to the revised bill?"
Japanese Prime Minister Kishida replied at the meeting: "The bill has been revised and has fully responded to the opinions and demands of (related parties)", and also cited as an example that the revision of the bill "reflects the voices of local governments and medical frontlines at the forefront."
Nakatani then released the Prime Minister's answer predicted by ChatGPT and displayed it on a display board, and praised the response given by AI as "more sincere and focused" than the Prime Minister's answer. Kishida Fumio retorted that "my answer is more detailed, and there are examples, which better reflect the actual situation." This defense caused a burst of laughter from the members present.
Prime Minister Kishida acknowledged, however, that conversational AI "has the potential to help administrators effectively use more information if used properly," and said he would consider how AI technology could be applied in the future.
This week's AI application recommendation
☑️ Perplexity AI : A powerful search tool that combines AI models with real-time search
☑️ Momo AI : AI plans your travel plan, iOS APP is available
☑️ Respell : Customize the AI work path, create a private AI workbench, and you can join the waiting list
☑️ Chatwebpage : AI reads webpage information with one click and starts a conversation
☑️ Genmo : Generate pictures or videos from text, which can be further edited
☑️ SpeechGPT : Let ChatGPT help you practice speaking
☑️ ChitGPT : A large language model specially designed by CSDN for developers
This week, a "joint letter" brought the topic of "super AI research and development and the development of human society" to the table again. Regardless of how the "joint letter" ends up, but under the torrent of the era of continuous advancement of AI technology, " "The relationship between AI and human beings" will be an issue that people need to face sooner or later.
Because it's there.