Artificial General Intelligence (AGI) News and Discussions

[wjfox]

SoftBank CEO says artificial general intelligence will come within 10 years

By Reuters
Published 2:49 AM EDT, Wed October 4, 2023

Tokyo Reuters — SoftBank CEO Masayoshi Son said he believes artificial general intelligence (AGI), artificial intelligence that surpasses human intelligence in almost all areas, will be realized within 10 years.

Speaking at the SoftBank World corporate conference, Son said he believes AGI will be ten times more intelligent than the sum total of all human intelligence. He noted the rapid progress in generative AI that he said has already exceeded human intelligence in certain areas.

“It is wrong to say that AI cannot be smarter than humans as it is created by humans,” he said. “AI is now self learning, self training, and self inferencing, just like human beings.”

Son has spoken of the potential of AGI — typically using the term “singularity” — to transform business and society for some years, but this is the first time he has given a timeline for its development.

He also introduced the idea of “Artificial Super Intelligence” at the conference which he claimed would be realized in 20 years and would surpass human intelligence by a factor of 10,000.

https://edition.cnn.com/2023/10/04/tech ... index.html

[Ozzie guy]

I’ve always understood AGI (Artificial General Intelligence) to be about general intelligence, not necessarily outperforming us in every domain.

Right now, there’s a referendum going on in Australia. It’s not the official voting date yet, but early voting is underway. However, the number of voting locations is limited, and finding where to vote can be a bit of a hassle.

Today, I used GPT-4, a publicly available AI, to find the closest voting location to my house. It wasn’t just a simple answer; GPT-4 browsed the web, hopping from one site to another, until it found the information.

Watching GPT-4 in action was like observing an autonomous agent. It navigated the digital realm, left websites when the info wasn’t there, and looked elsewhere. It was quite the sight.

And this isn't all GPT-4 can do; it now has the ability to understand images among other things.

Imagine seeing this AI back in 2013. What would we have called it?

Oh, and by the way, this post was crafted by GPT-4. I’m Ozzie Guy, (This last paragraph is still GPT-4 actually I also had it browse a couple of my posts on this site) but today GPT-4 is stepping in as I’m a bit out of it due to some sleeping meds. I did have to nudge it a bit to make this post more human-like, but I think it did a decent job. I’ll be uploading some snapshots of our earlier conversation about finding the voting location, with sensitive bits blurred out.

[wjfox]

Moved to news and current events.

[FuturismFan]

In my opinion, once the Coffee Test has been passed, AGI will have finally arrived.

https://en.wikipedia.org/wiki/Artificia ... telligence

A machine is required to enter an average American home and figure out how to make coffee: find the coffee machine, find the coffee, add water, find a mug, and brew the coffee by pushing the proper buttons. This has not yet been completed.

[Cyber_Rebel]

The Coffee Test is an interesting one, so I'd have to wonder something: Could GPT-4 with its vision capabilities, and embodiment be fully capable of accomplishing it?

This was back in March, when we knew GPT-4 had multimodal capability but Open AI wished to hold it back for refinement:



Check out this blog for the use case examples regarding GPT-Vision: https://blog.roboflow.com/gpt-4-vision/

Key thing noted here is "object detection," which itself could be solved by the implementation of embodiment in a robotic shell, especially if we take the recent advances of Google's RT2-X into account.

RT-2-X demonstrates understanding of spatial relationships between objects.
RT-2-X demonstrates skills that the RT-2 model was not capable of previously, including better spatial understanding. For example, if we ask the robot to "move apple near cloth" instead of "move apple on cloth" the trajectories are quite different. By changing the preposition from "near" to "on", we can modulate the actions that robot takes.

RT-2-X shows that combining data from other robots into the training improves the range of tasks that can be performed even by a robot that already has large amounts of data available – but only when utilizing a sufficiently high-capacity architecture.

Now, assuming Open AI takes a similar approach in regard to their embodiment model Neo-X: OpenAI’s Leap into Robotics: Introducing Neo

The vision is that, as Neo moves and interacts with its environment, it grows smarter, adapting to a plethora of applications, much like humans.

Balanced Performance: The promise is that Neo will move much like humans. Engineers have crafted its limbs to mirror human muscle movement, allowing it to navigate our spaces with ease.

-----------------------------------------------------------------------------

Implementing all these advances into one synthetic unit, GPT-Vison, voice capabilities, and mobile embodiment with spatial awareness and the capability to learn from prior examples of trained datasets, and I believe that the Coffee Test actually could be passed, possibly as soon as next year depending on when Open AI (via 1X) releases their units.

The real question is, will our goalpost move yet again? Or will we finally recognize this as an actual AGI?

[raklian]

The Coffee Test is an interesting one, so I'd have to wonder something: Could GPT-4 with its vision capabilities, and embodiment be fully capable of accomplishing it?

Implementing all these advances into one synthetic unit, GPT-Vison, voice capabilities, and mobile embodiment with spatial awareness and the capability to learn from prior examples of trained datasets, and I believe that the Coffee Test actually could be passed, possibly as soon as next year depending on when Open AI (via 1X) releases their units.

The real question is, will our goalpost move yet again? Or will we finally recognize this as an actual AGI?

I do think we have all of the prerequisite technologies to make it happen. Someone just needs to figure out the best configuration to put all of them together.

[Cyber_Rebel]

Artificial General Intelligence Is Already Here

Today’s most advanced AI models have many flaws, but decades from now, they will be recognized as the first true examples of artificial general intelligence.



BY BLAISE AGÜERA Y ARCAS AND PETER NORVIG
OCTOBER 10, 2023

(Noemamag)

Artificial General Intelligence (AGI) means many different things to different people, but the most important parts of it have already been achieved by the current generation of advanced AI large language models such as ChatGPT, Bard, LLaMA and Claude. These “frontier models” have many flaws: They hallucinate scholarly citations and court cases, perpetuate biases from their training data and make simple arithmetic mistakes. Fixing every flaw (including those often exhibited by humans) would involve building an artificial superintelligence, which is a whole other project.

Nevertheless, today’s frontier models perform competently even on novel tasks they were not trained for, crossing a threshold that previous generations of AI and supervised deep learning systems never managed. Decades from now, they will be recognized as the first true examples of AGI, just as the 1945 ENIAC is now recognized as the first true general-purpose electronic computer.

The ENIAC could be programmed with sequential, looping and conditional instructions, giving it a general-purpose applicability that its predecessors, such as the Differential Analyzer, lacked. Today’s computers far exceed ENIAC’s speed, memory, reliability and ease of use, and in the same way, tomorrow’s frontier AI will improve on today’s.

But the key property of generality? It has already been achieved.

[wjfox]

[raklian]

What's even more interesting is what happens after 2025 or after 2030 if we're conservative. The scale of AGI growth after that is going to blow minds.

AGI will be a LLM on steroids with a much larger parameter count (in the realm of tens of trillions). There are those who argue any LLM will never become an AGI, though. Still, I argue there will be different types of AGI.

The difficulty AGI encounters in solving the problem of aging or any intractable problem gets exponentially reduced every 1-2 years from that point on. We are about to enter a world that is utterly unrecognizable from what we've experienced.

[Tadasuke]

My answer for all that AI talk is this really depressing video:



We need to improve humans and this world before creating potential superintelligence. However, I believe that AI overall isn't progressing anywhere nearly as quickly or well as some claim (especially on the Singularity reddit and associated discords).

[spryfusion]

Timestamps:
(0:00:00) - Measuring AGI
(0:11:41) - Do we need new architectures?
(0:16:26) - Is search needed for creativity?
(0:19:19) - Superhuman alignment
(0:29:58) - Impact of DeepMind on safety vs capabilities
(0:34:03) - Timelines
(0:41:24) - Multimodality

[funkervogt]

Timestamps:
(0:00:00) - Measuring AGI
(0:11:41) - Do we need new architectures?
(0:16:26) - Is search needed for creativity?
(0:19:19) - Superhuman alignment
(0:29:58) - Impact of DeepMind on safety vs capabilities
(0:34:03) - Timelines
(0:41:24) - Multimodality

Interesting that he says The Age of Spiritual Machines influenced his own predictions about AGI.

[raklian]

Google AI Chief Says There’s a 50% Chance We’ll Hit AGI in Just 5 Years
by NOOR AL-SIBAI

In an interview with tech podcaster Dwarkesh Patel, DeepMind co-founder Shane Legg said that he still thinks that researchers have a 50-50 chance of achieving artificial general intelligence (AGI), a stance he publicly announced at the very end of 2011 on his blog.

Legg apparently began looking towards his 2028 goalpost all the way back in 2001 after reading "The Age of Spiritual Machines," the groundbreaking 1999 book by fellow Google AI luminary Ray Kurzweil that predicts a future of superhuman AIs.

"There were two really important points in his book that I came to believe as true," he explained. "One is that computational power would grow exponentially for at least a few decades. And that the quantity of data in the world would grow exponentially for a few decades."

Paired with an understanding of the trends of the era, such as the deep learning method of teaching algorithms to "think" and process data the way human brains do, Legg wrote back at the start of the last decade that in the coming ones, AGI could well be achieved — so long as "nothing crazy happens like a nuclear war."

https://futurism.com/google-deepmind-agi-5-years

[wjfox]

[spryfusion]

Google DeepMind paper: "Levels of AGI: Operationalizing Progress on the Path to AGI"

Abstract:

We propose a framework for classifying the capabilities and behavior of Artificial General Intelligence (AGI) models and their precursors. This framework introduces levels of AGI performance, generality, and autonomy. It is our hope that this framework will be useful in an analogous way to the levels of autonomous driving, by providing a common language to compare models, assess risks, and measure progress along the path to AGI. To develop our framework, we analyze existing definitions of AGI, and distill six principles that a useful ontology for AGI should satisfy. These principles include focusing on capabilities rather than mechanisms; separately evaluating generality and performance; and defining stages along the path toward AGI, rather than focusing on the endpoint. With these principles in mind, we propose “Levels of AGI” based on depth (performance) and breadth (generality) of capabilities, and reflect on how current systems fit into this ontology. We discuss the challenging requirements for future benchmarks that quantify the behavior and capabilities of AGI models against these levels. Finally, we discuss how these levels of AGI interact with deployment considerations such as autonomy and risk, and emphasize the importance of carefully selecting Human-AI Interaction paradigms for responsible and safe deployment of highly capable AI systems.

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

[wjfox]

Sam Altman on AI next year: "The model capability will have taken such a leap forward that no one expected... that much progress"

[wjfox]