Tech ChatGPT developer intends to establish office in Japan

[thomas]

Today, Sam Altman, CEO of OpenAI, met Prime Minister Fumio Kishida and other senior officials in Tokyo, pledging to work with the government to protect user privacy and safety. Altman told reporters that he explained to Kishida, who will preside over the G-7 summit in his constituency of Hiroshima in May, about the pros and cons of ChatGPT while saying the premier showed an interest in the AI technology. The visit to Japan seems to have been arranged by the ruling party, the LDP.

In the afternoon, Altman attended a gathering of ruling Liberal Democratic Party (LDP) lawmakers for discussions. At the start of meeting, he said: "The development of AI will be one of the most important technological revolutions ever," adding that governments have a role to play in protecting their citizens when new technologies emerge. He also said his company was willing to cooperate with Japan and other countries to do this.

The LDP's draft proposal calls for the "thorough utilization by the national government," employing it on a trial basis, for example, in answers to parliamentary questions. According to Altman, more than 1 million people use ChatGPT every day in Japan. "What's happening in Japan is just so important, both from a technological adoption standpoint but also a policy perspective," Altman said at the meeting with the LDP. By visiting Japan, which is open to the use of ChatGPT, Altman appears to be trying to broaden support for OpenAI's initiative. He also hinted at the idea of establishing an office in Japan in the near future.

And the downsides, some of which we have already experienced with our Jaybot, too:

Dealing with the creation and dissemination of false information is another major challenge for generative AI. While ChatGPT can respond to questions in natural, humanlike sentences, it can also create plausible statements with factual errors based on incorrect training data. If used maliciously, AI tools could lead to widespread dissemination of false information, potentially creating social instability. Although Japan has not had much regulatory debate on the technology so far, the LDP's draft proposal also raises the "risk of undue intervention [by AI tools] in the democratic process" as a point of contention.

ChatGPT developer eager to open office in Japan amid concern over use

A developer of ChatGPT said on April 10 that his company plans to open an office in Japan amid growing concerns about the chatbot's unauthorized collection of personal data and impact on learning environments.

english.kyodonews.net

Paywall alert:

OpenAI CEO vows to work with Japan on user protections

Sam Altman meets Kishida, LDP officials, hints at local ChatGPT office

asia.nikkei.com

 

[musicisgood]

Actually I like this AI stuff is pretty cool really especially like if you're in a hurry and you want some immediate the answers and the chances are the answers are very helpful whether the correct or not that's debatable but from my experience so far they seem to be pretty helpful information coming out of this new technology I actually I don't think it's new but it's just something different that is been updated from I guess Google search.

 

[bentenmusume]

Today, Sam Altman, CEO of OpenAI, met Prime Minister Fumio Kishida and other senior officials in Tokyo, pledging to work with the government to protect user privacy and safety. Altman told reporters that he explained to Kishida, who will preside over the G-7 summit in his constituency of Hiroshima in May, about the pros and cons of ChatGPT while saying the premier showed an interest in the AI technology. The visit to Japan seems to have been arranged by the ruling party, the LDP.

Oh, it's wonderful to know that the CEO of OpenAI and the leader of the LDP -- two entities famously known for their commitment to ethics and the welfare of the general populace -- are teaming up on this one. With their efforts combined, there's surely no way that ChatGPT and its presence in Japan can be anything but a bastion of transparency and a boon to society unsullied by scandal or any concern of misuse or wrongdoing.

Actually I like this AI stuff is pretty cool really especially like if you're in a hurry and you want some immediate the answers and the chances are the answers are very helpful whether the correct or not that's debatable but from my experience so far they seem to be pretty helpful information coming out of this new technology

First off, I'm sorry for exploding at you the other day. I wasn't really in a good place, but that's no excuse for my being so mean-spirited and hostile toward you. So again, my apologies.

That said, the bolded part is exactly the problem with "this AI stuff" and with modern society in general.

If the answers aren't correct, how can they be "helpful"?
And if you're searching for information you don't already know, then how do you know if they're correct or not?

I actually I don't think it's new but it's just something different that is been updated from I guess Google search.

Believe it or not, I actually agree with you here.

I don't think it's all that different at all -- it basically just spits out the same stuff you could find by searching for things yourself, except with the sources obfuscated (yes, mdchachi, I know that Bing AI does reveal its sources -- often to hilarious effect, as it links to completely unrelated sites that fail to substantial its fabrications or inaccuracies) and presented in authoritative language that can all-too-easily lead gullible, uneducated people to think they're being enlightened when they're actually just being fed a heaping dose of bullshit.

 

[Lothor]

Although I agree with you about your concerns, particularly about LDP and ChatGPT teaming up, I have found ChatGPT helpful in two contexts in my work.

First, I had to explain a sentence in English to a Japanese translator who is translating a book into Japanese. I 'knew' what the sentence meant but couldn't find the words to explain it. It gave me a very good explanation that I agreed with, which I gave to the translator.

Second, I was proofreading an academic paper and the abbreviation AIC came up but was not defined. Some of my later changes required me to know what it stood for. I copied the sentences into ChatGPT with the question 'What do you think AIC stands for?' and it gave me an answer that fitted the context perfectly. When it came out with 'Akaike's information criterion', I had an 'Of course!' moment.

In both cases, I had to make an informed judgement - the main problem seems to be an uninformed reliance on AI.

 

[mdchachi]

Actually I like this AI stuff is pretty cool really especially like if you're in a hurry and you want some immediate the answers and the chances are the answers are very helpful whether the correct or not that's debatable but from my experience so far they seem to be pretty helpful information coming out of this new technology I actually I don't think it's new but it's just something different that is been updated from I guess Google search.

I've been using it (bing ai) quite a bit myself as an alternative to traditional search.
I recently discovered it can get through paywalls. So I can ask it to summarize a specific news article since the articles are open to bots for indexing.

mdchachi, I know that Bing AI does reveal its sources -- often to hilarious effect, as it links to completely unrelated sites that fail to substantial its fabrications or inaccuracies) and presented in authoritative language that can all-too-easily lead gullible, uneducated people to think they're being enlightened when they're actually just being fed a heaping dose of bullshit.

I'm pretty sure Bing is faking it. I think it's probably feeding its traditional search results to chatGPT and then guessing what to footnote. It doesn't know for sure what chatGPT is going to come up with or if it was sourced from its links.

If the answers aren't correct, how can they be "helpful"?
And if you're searching for information you don't already know, then how do you know if they're correct or not?

I just use the smell test. If I ask somebody on reddit what the ルビコン川 is, I can't be sure they know what they are talking about. But if it sounds reasonable I'll probably believe it. The response Jaybot gave me moments ago seemed reasonable. So I'll assume it's correct for now and move on. (Incidentally the term came up just now as somebody surmised we'd already crossed the ルビコン川 on tv as they were talking about chatGPT so I was wondering if it meant the same thing as "uncanny valley"). Unlike Reddit, a response is guaranteed and immediate.

 

[bentenmusume]

First, I had to explain a sentence in English to a Japanese translator who is translating a book into Japanese. I 'knew' what the sentence meant but couldn't find the words to explain it. It gave me a very good explanation that I agreed with, which I gave to the translator.

Fair enough. In my case, I trust my own knowledge and my ability to explain things in Japanese more than I trust an ignorant AI chatbot that doesn't really understand anything, so I choose not to avail myself of such resources.

Second, I was proofreading an academic paper and the abbreviation AIC came up but was not defined. Some of my later changes required me to know what it stood for. I copied the sentences into ChatGPT with the question 'What do you think AIC stands for?' and it gave me an answer that fitted the context perfectly. When it came out with 'Akaike's information criterion', I had an 'Of course!' moment.

Again, wonderful. In that case, I would probably just Google it and evaluate the resources for myself. I don't really see what ChatGPT offers beyond doing one's own research, but if you felt it helpful, I guess in some way I can respect that. Still, I would rather do the research myself and speak in my own words.

I just use the smell test. If I ask somebody on reddit what the ルビコン川 is, I can't be sure they know what they are talking about. But if it sounds reasonable I'll probably believe it. The response Jaybot gave me moments ago seemed reasonable. So I'll assume it's correct for now and move on. (Incidentally the term came up just now as somebody surmised we'd already crossed the ルビコン川 on tv as they were talking about chatGPT so I was wondering if it meant the same thing as "uncanny valley"). Unlike Reddit, a response is guaranteed and immediate.

I hate this logic. "I don't know if humans are right or not, so I might as well ask a Chatbot."

What does "sounds reasonable" mean? If you don't know the answer to something, anything "sounds reasonable." Both you and Thomas -- two people who I trust as very enlightened and intelligent individuals -- flagged Jaybot's response to the question about Tokyo transportation systems where it made up literal fictional nonsense about the Inokashira Line being expanded to Kichijoji in 2013 as "informative", despite the fact that this is so far away from anything that could be considered the truth that it is either utterly hilarious or utterly pathetic, depending on how you choose to see it.

If you can only trust ChatGPT to answer stuff you already kinda-sorta know the answers to, then what is even the point? If you're trusting ChatGPT for stuff you don't really know, then -- even more so -- what is the point at all, since it could literally just be feeding you absolute nonsense.

"A response is guaranteed and immediate." Great! So BS and misinformation can be spread at the speed of light. Truly the next evolution of internet discourse.

/barf