Google Pathways: A multipurpose proto-AGI in development

Post Reply
User avatar
Yuli Ban
Posts: 4631
Joined: Sun May 16, 2021 4:44 pm

Google Pathways: A multipurpose proto-AGI in development

Post by Yuli Ban »

Really fascinating stuff!!

Google is developing a nimble, multi-purpose AI that can perform millions of tasks. Called Pathways, Google’s solution seeks to centralize disparate AI into one powerful, all-knowing algorithm, Jeff Dean, Google’s AI chief revealed
Artificial intelligence is already capable of doing some incredibly useful things, like predicting flooding, diagnosing disease, and instantly translating languages. Advances in neural networks coupled with enormous computational power have allowed tech companies to create incrementally smarter AI models over the last decade.

Jeff Dean, Google’s AI chief, thinks we’re just scratching the surface. Speaking at the TED conference in Monterey, California, this week, he revealed that Google is developing a nimble, multi-purpose AI that can perform millions of tasks. Called Pathways, Google’s solution seeks to centralize disparate AI into one powerful, all-knowing algorithm.

“We’re pretty excited about this,” Dean said. “These kinds of general purpose intelligence systems that have a deeper understanding of the world will really enable us to tackle some of the greatest problems that humanity faces.”

For example, he said, “[w]e’ll be able to engineer better medicines by infusing these models with knowledge of chemistry and physics; we’ll be able to advance educational systems by providing more individualized tutoring; we’ll be able to tackle really complicated issues like climate change, perhaps engineering of clean energy solutions.”
This seems to be more of a teaser than any sort of announcement, but I'll definitely stay tuned.
And remember my friend, future events such as these will affect you in the future
User avatar
Ozzie guy
Posts: 486
Joined: Sun May 16, 2021 4:40 pm

Re: Google Pathways: A multipurpose proto-AGI in development

Post by Ozzie guy »

I am unable to read the article when is this expected to be made could we be looking at having proto AGI by the end of the year!!??
(I remember sounding like an insane delusional optimist near the start of the year when I was saying someone I trusted said proto AGI is possible this year)

It feels like we are having a GPT3 level breakthrough every couple of months!!

What is the goal after proto AGI?
I originally assumed the next goal was human level AGI however recently I have been seeing an idea of AGI that is not Proto but not human level.
jamestiago
Posts: 18
Joined: Thu Aug 12, 2021 3:04 pm

Re: Google Pathways: A multipurpose proto-AGI in development

Post by jamestiago »

At this point it's hard to tell what tomorrow holds for mankind.
User avatar
Ozzie guy
Posts: 486
Joined: Sun May 16, 2021 4:40 pm

Google confirmed to be focused on making Proto AGI

Post by Ozzie guy »

I went digging a bit and found more information on Googles pathways.
https://blog.ted.com/tech-comeback-note ... dmonterey/

What Jeff Dean head of googles AI project wants is literally exactly inline with Yuli's/Starspawn0's definition of Proto AGI. He has also confirmed that Google' current project is moving in the exact direction of Proto AGI or is literally Proto AGI itself.

"Jeff Dean, computer systems developer

Big idea: We’ve made tremendous progress in neural networks and computational power over the past two decades. But to achieve the full power of AI, we need to fix three key things.

How? Jeff Dean, the head of Google’s AI efforts, has been embedded in the world of artificial intelligence for decades. Despite the exponential evolution of neural networks and computational power, he thinks there are three key areas of focus still in need of fixing to realize the true potential of machine learning. First off: multi-task models. Whereas current AI systems are trained for specific tasks and learn new tasks from scratch each time — a multi-task model enables a system to do thousands of tasks and leverage that expertise to complete totally new tasks. Likewise, Dean believes we should train AI across images, text, sound and video simultaneously — rather than just one-by-one, as current systems do. Lastly, he advocates for “sparse models” (instead of current dense models, which activate a whole system for each task) that would only activate relevant parts for a given task (much like the human brain). Drawing on these ideas, Dean reveals publicly for the first time the next phase in Google’s AI plans: a new system called Pathways, which aims to generalize across millions of tasks and take a major step forward in how we build powerful, responsible artificial intelligence."


I felt like this is big enough news to deserve it's own thread, partially because I want people to see this I don't want this news to turn into a post reply that is ignored.
User avatar
raklian
Posts: 1746
Joined: Sun May 16, 2021 4:46 pm
Location: North Carolina

Re: Google confirmed to be focused on making Proto AGI

Post by raklian »

Set and Meet Goals wrote: Sat Aug 14, 2021 4:26 am
I felt like this is big enough news to deserve it's own thread, partially because I want people to see this I don't want this news to turn into a post reply that is ignored.
Merged it with this thread. There is no point having multiple threads about the same topic.
To know is essentially the same as not knowing. The only thing that occurs is the rearrangement of atoms in your brain.
User avatar
Yuli Ban
Posts: 4631
Joined: Sun May 16, 2021 4:44 pm

Re: Google confirmed to be focused on making Proto AGI

Post by Yuli Ban »

Set and Meet Goals wrote: Sat Aug 14, 2021 4:26 am I went digging a bit and found more information on Googles pathways.
https://blog.ted.com/tech-comeback-note ... dmonterey/

What Jeff Dean head of googles AI project wants is literally exactly inline with Yuli's/Starspawn0's definition of Proto AGI. He has also confirmed that Google' current project is moving in the exact direction of Proto AGI or is literally Proto AGI itself.

"Jeff Dean, computer systems developer

Big idea: We’ve made tremendous progress in neural networks and computational power over the past two decades. But to achieve the full power of AI, we need to fix three key things.

How? Jeff Dean, the head of Google’s AI efforts, has been embedded in the world of artificial intelligence for decades. Despite the exponential evolution of neural networks and computational power, he thinks there are three key areas of focus still in need of fixing to realize the true potential of machine learning. First off: multi-task models. Whereas current AI systems are trained for specific tasks and learn new tasks from scratch each time — a multi-task model enables a system to do thousands of tasks and leverage that expertise to complete totally new tasks. Likewise, Dean believes we should train AI across images, text, sound and video simultaneously — rather than just one-by-one, as current systems do. Lastly, he advocates for “sparse models” (instead of current dense models, which activate a whole system for each task) that would only activate relevant parts for a given task (much like the human brain). Drawing on these ideas, Dean reveals publicly for the first time the next phase in Google’s AI plans: a new system called Pathways, which aims to generalize across millions of tasks and take a major step forward in how we build powerful, responsible artificial intelligence."


I felt like this is big enough news to deserve it's own thread, partially because I want people to see this I don't want this news to turn into a post reply that is ignored.
We're going to need to see benchmarks and tests, but if they can accomplish even half of what this TED talk promised, we'll be in for a VERY wild ride VERY soon.
And remember my friend, future events such as these will affect you in the future
User avatar
Cyber_Rebel
Posts: 331
Joined: Sat Aug 14, 2021 10:59 pm
Location: New Dystopios

Re: Google Pathways: A multipurpose proto-AGI in development

Post by Cyber_Rebel »

Image

At last, some great news has arrived. Literally, this is one of the last hopes for humanity I currently have, the potential for better harnessing the knowledge and potential in society is almost limitless when you unironically remove flawed variables from the equation. The data it could process would ideally improve upon itself multitudes times faster than any group of people or think tanks, and without the added "benefit" of biases, special interest, or conflict of interest which comes with our biological shortcomings.

Finally, this feels like a real step into the future. Even if this is premature conjecture, the ideal comes closer than anything so far. Aside from disease and climate models, new technological blueprints we couldn't think of might be achievable. To say nothing of central governance should this be the legit genuine article we hope it is.

Just keep it away from 4chan or Reddit, and I think we'll be alright. The data it's fed is a concern, but maybe it'll filter out the dross ala MGS "Societal Sanity" style. Hello again Future Timeline, hopefully we have good things on the horizon for once.
Last edited by Cyber_Rebel on Sat Aug 14, 2021 11:36 pm, edited 1 time in total.
User avatar
Ozzie guy
Posts: 486
Joined: Sun May 16, 2021 4:40 pm

Re: Google Pathways: A multipurpose proto-AGI in development

Post by Ozzie guy »

Google is Actively building an AGI that can learn an Infinite number of Tasks.

Googles AGI will use all forms of Modality possible.
Image

It will be able to learn an infinite number of tasks generalising its current knowledge to each new task it is trying to learn. (e.g. in the first photo adding task number 1001)
Image

Image

They want to move to general propose intelligent systems with a deeper understanding of the world that can solve all of humanities problems.
Image

Image

Google has hundreds of researches employed in AI safety.
User avatar
TanishaTanTan
Posts: 21
Joined: Tue Jan 10, 2023 10:23 pm

Re: Google Pathways: A multipurpose proto-AGI in development

Post by TanishaTanTan »

Yuli Ban wrote: Thu Aug 12, 2021 2:01 am Really fascinating stuff!!

Google is developing a nimble, multi-purpose AI that can perform millions of tasks. Called Pathways, Google’s solution seeks to centralize disparate AI into one powerful, all-knowing algorithm, Jeff Dean, Google’s AI chief revealed
Artificial intelligence is already capable of doing some incredibly useful things, like predicting flooding, diagnosing disease, and instantly translating languages. Advances in neural networks coupled with enormous computational power have allowed tech companies to create incrementally smarter AI models over the last decade.

Jeff Dean, Google’s AI chief, thinks we’re just scratching the surface. Speaking at the TED conference in Monterey, California, this week, he revealed that Google is developing a nimble, multi-purpose AI that can perform millions of tasks. Called Pathways, Google’s solution seeks to centralize disparate AI into one powerful, all-knowing algorithm.

“We’re pretty excited about this,” Dean said. “These kinds of general purpose intelligence systems that have a deeper understanding of the world will really enable us to tackle some of the greatest problems that humanity faces.”

For example, he said, “[w]e’ll be able to engineer better medicines by infusing these models with knowledge of chemistry and physics; we’ll be able to advance educational systems by providing more individualized tutoring; we’ll be able to tackle really complicated issues like climate change, perhaps engineering of clean energy solutions.”
This seems to be more of a teaser than any sort of announcement, but I'll definitely stay tuned.
It is interesting
User avatar
TanishaTanTan
Posts: 21
Joined: Tue Jan 10, 2023 10:23 pm

Re: Google Pathways: A multipurpose proto-AGI in development

Post by TanishaTanTan »

Ozzie guy wrote: Thu Aug 12, 2021 3:30 am I am unable to read the article when is this expected to be made could we be looking at having proto AGI by the end of the year!!??
(I remember sounding like an insane delusional optimist near the start of the year when I was saying someone I trusted said proto AGI is possible this year)

It feels like we are having a GPT3 level breakthrough every couple of months!!

What is the goal after proto AGI?
I originally assumed the next goal was human level AGI however recently I have been seeing an idea of AGI that is not Proto but not human level.
I know that developing AGI is a very complex process. I'll keep an eye on this development. I also recently read about development services, found https://dashdevs.com/ information about it here. I myself would someday like to participate in the creation and development of any global project. I will do my best to achieve this. I plan to take some courses for additional knowledge.
In general, this information scares me a little... I'm afraid of the future
Post Reply