Artificial General Intelligence (AGI) News and Discussions

[Yuli Ban]

[Yuli Ban]

[wjfox]

[agi]

deleted

[wjfox]

[weatheriscool]

ChatGPT may be the fastest-growing consumer app in internet history, reaching 100 million users in just over 2 months, UBS report says
Sawdah Bhaimiya
Feb 2, 2023, 3:45 AM

ChatGPT may be the fastest-growing app in history with 100 million users after just over two months.
A UBS report showed that it took TikTok over nine months to reach the same number of people.
OpenAI is rolling out a $20-a-month subscription to ChatGPT for easy access and faster responses.

ChatGPT may have become the fastest-growing app in history after it was estimated to have reached 100 million monthly active users, or MAUs, by the end of January, a report by Swiss banking giant UBS said.

The OpenAI-owned chatbot has surpassed 100 million MAUs just over two months after its launch — comfortably beating the growth speeds of major apps like TikTok and Instagram, according to the UBS note, which Insider has viewed.

https://www.businessinsider.com/chatgpt ... udy-2023-2

[agi]

deleted

[Yuli Ban]

I've begun strongly shifting my timescales.

I think now that pseudo/frozen AGI is imminent, possibly as soon as this year.

But full/true/continuously learning AGI? Bad news. I don't see that arising for many, many years. Listening to actual engineers and researchers, they agree that something a bit like frozen AGI might be possible if Gato can be scaled up... but a sentient, human-level AGI system? That's not it, and we might still be 10+ years away, even now.

Indeed, I'm starting to get the sense that this whole boom may in fact be a false start towards AGI, that language models have fooled us because of the nature of language itself.

We'll see if this is the correct viewpoint in a few years. I think by 2025-2026, we'll be able to see for sure whether AGI is imminent or not. But I'm definitely starting to shift away from that viewpoint.


Should probably be stressed that my timelines lengthening isn't an example of increased pessimism; more than I was almost certainly too optimistic to begin with because I was seeing only what I wanted to see and disregarded any information to the contrary.

"Generalized systems are within reach, but capable general AI is 10+ years away at least and there's no known way to get there; even large language models are not a good path to it, and here's why:" — immediately discarded! You're failing to account for exponential growth! Law of accelerating returns! AGI will be here by 2025, just watch!

"No, it's not; I literally work with deep neural networks; these are intractable problems that we have no—" This pop-sci article about ChatGPT showing theory of mind and Arxiv paper on chain of memory learning (that I can't understand) disprove you, we're much further along!

"Literally every machine learning researcher worth their salt predicted exactly these sorts of scaling effects in the 80s and 90s; they just couldn't test them because—" Did you hear that literally no one on Earth can understanding why Bing Chat can do what it does? I bet AGI ought to be here even sooner than expected if multiple different methodologies that I don't understand are crossed!



Yeah, no more of that.

Plain, cold fact is that, for way too long, I fell victim to this thinking:

[Nero]

That is sort of why I have stated previously that I would not be confident in any prediction beyond the short term for AI technology, the biggest reason is that the largest advances that happen are rarely predictable, the progress in the field has been gigantic yes but it's also very difficult to quantify. It is entirely possible that by the mid 2020's AGI is available for the masses to use, it is equally possible that there is not sufficient enough knowledge to create one.

The largest issue really is that the future of AI is not as clear as the future of electronics or computing which can be accurately predicted over the course of more than a year. That is definitely not the case with AI. The only things that are more or less certain are that there will be huge advances before the next year and those advances will undeniably change many of the predictions made before and after, the capabilities of this technology are unknown and they are advancing in a way that we are effectively flying blind.

https://developer.nvidia.com/blog/desig ... -learning/

"The latest NVIDIA Hopper GPU architecture has nearly 13,000 instances of AI-designed circuits."

This is the biggest black swan so to speak, the way in which these incredible new technologies can create and design further technology is a huge unknown. AI inevitably leads to more AI. We are not at the stage where any accurate prediction of what this will make possible is understood or even possible to conceptualize with such limited information. Imagine trying to predict what modern social media would look like when all we had were massive room-sized computers. That is where we are with AI but accelerating at a much faster pace with no real way of predicting the usage of the technology or what it would lead to.

The only real certainty is that it is best to be uncertain about where this technology will go even in the short term of beyond the next year. Your predictions, and everyone else's will inevitably fail to account for something that will only be clear with hindsight.

[ººº]

I think now that pseudo/frozen AGI is imminent, possibly as soon as this year.

But full/true/continuously learning AGI? Bad news. I don't see that arising for many, many years. Listening to actual engineers and researchers, they agree that something a bit like frozen AGI might be possible if Gato can be scaled up... but a sentient, human-level AGI system? That's not it, and we might still be 10+ years away, even now.

I consider any date before 2050 as quite soon.

[Ozzie guy]

Indeed, I'm starting to get the sense that this whole boom may in fact be a false start towards AGI, that language models have fooled us because of the nature of language itself.

I never really took LLM seriously and I think subconsciously expected this boom and bust of hype. On this website I have sounded like a schizo for using the word GATO in every AI related post I made. This is because it made no sense to me that people followed LLM I kept thinking yes this is a cool tool but GATO is actually a proto type of proto ( ) AGI.

In a similar vein why should we focus on what Sam Altman is saying all OpenAI have bult is impressive language models when Deep Mind is likely building some kind of proto/frozen AGI.

I think the best way of looking at things is at any given time to look at the one AI that is the closest to AGI and then predict the AI being built that will supplant it as closest to AGI. It seems impossible to look into the future further than that although I really want to as a coping mechanism.

[spryfusion]

[wjfox]

Nvidia CEO: "We're going to accelerate AI by another million times"

28th February 2023

In a recent earnings call, the boss of Nvidia Corporation, Jensen Huang, outlined his company's achievements over the last 10 years and predicted what might be possible in the next decade.

https://www.futuretimeline.net/blog/202 ... s-2030.htm

[wjfox]

[wjfox]

Meta’s powerful AI language model has leaked online — what happens now?

Mar 8, 2023, 1:15 PM GMT

Two weeks ago, Meta announced its latest AI language model: LLaMA. Though not accessible to the public like OpenAI’s ChatGPT or Microsoft’s Bing, LLaMA is Meta’s contribution to a surge in AI language tech that promises new ways to interact with our computers as well as new dangers.

Meta did not release LLaMA as a public chatbot (though the Facebook owner is building those too) but as an open-source package that anyone in the AI community can request access to. The intention, said the company, is “further democratizing access” to AI to spur research into its problems. Meta benefits if these systems are less buggy, so will happily spend the money to create the model and distribute it for others to troubleshoot with.

“Even with all the recent advancements in large language models, full research access to them remains limited because of the resources that are required to train and run such large models,” said the company in a blog post. “This restricted access has limited researchers’ ability to understand how and why these large language models work, hindering progress on efforts to improve their robustness and mitigate known issues, such as bias, toxicity, and the potential for generating misinformation.”

However, just one week after Meta started fielding requests to access LLaMA, the model was leaked online. On March 3rd, a downloadable torrent of the system was posted on 4chan and has since spread across various AI communities, sparking debate about the proper way to share cutting-edge research in a time of rapid technological change.

Some say the leak will have troubling consequences and blame Meta for distributing the technology too freely. “Get ready for loads of personalized spam and phishing attempts,” tweeted cybersecurity researcher Jeffrey Ladish after the news broke. “Open sourcing these models was a terrible idea.”

https://www.theverge.com/2023/3/8/23629 ... ine-misuse

[wjfox]

[wjfox]

[wjfox]

Bard: Google's rival to ChatGPT launches for over-18s

1 hour ago

Google will begin rolling out its AI chatbot Bard today, but it will only be available to certain users and they will have to be over the age of 18.

Unlike its viral rival ChatGPT, it can access up-to-date information from the internet and has a "Google it" button which accesses search.

It also namechecks its sources for facts, such as Wikipedia.

But Google warned Bard would have "limitations" and said it might share misinformation and display bias.

This is because it "learns" from real-world information, in which those biases currently exist - meaning it is possible for stereotypes and false information to show up in its responses.

https://www.bbc.com/news/technology-65018107

[wjfox]

[wjfox]

Bill Gates: AI is most important tech advance in decades

1 hour ago

Microsoft co-founder Bill Gates says the development of artificial intelligence (AI) is the most important technological advance in decades.

In a blog post on Tuesday, he called it as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone.

"It will change the way people work, learn, travel, get health care, and communicate with each other," he said.

He was writing about the technology used by tools such as chatbot ChatGPT.

Developed by OpenAI, ChatGPT is an AI chatbot which is programmed to answer questions online using natural, human-like language.

https://www.bbc.com/news/technology-65032848