How Close Are We Really to Achieving AGI? AI Expert Says 49%

In today's episode, we're charting the course of Artificial General Intelligence (AGI) as of June 2023, guided by the timeline presented by AI expert Dr. Alan D. Thompson.

A critical turning point came in 2017, with the birth of the attention mechanism by Google researchers, paving the way for the creation of Transformers. These are the foundation of powerful AI models like GPT, and according to Dr. Thompson, we're about halfway to achieving AGI from here.

The period between 2017 and 2023 has been rich with major AI advancements, each one a stepping stone on our path to AGI. These include our deeper understanding of GPT-3's capabilities, innovative applications of AI in microchip design, and the emergence of multitasking models like DeepMind's Gato.

By 2020, we had made roughly 30% progress towards AGI, nudging up to 31% by 2021. Significant leaps, like the unveiling of GPT-3's remarkable abilities and the use of AI in hardware design, were clear indicators of this progress. In May 2022, the introduction of DeepMind's Gato model propelled us to 39% completion, thanks to its impressive capabilities and the demonstration of human-like general intelligence.

As 2022 came to a close, progress continued with developments in autonomous AI systems and AI-assisted human tasks. Into 2023, we saw AGI estimates rise to 41%, largely due to Microsoft's innovative use of ChatGPT in operating robotic components. The introduction of Google's PaLM-E 562B, capable of planning its own future based on visual and language data, took AGI estimates up to 42%.

The launch of the even more powerful GPT-4 model and the integration of AI into robotics have substantially fast-tracked our journey towards AGI. Stay tuned to "Curious Future" for more updates on the fascinating world of AI, and don't forget to subscribe for the latest in AI, technology, and innovation.

CURIOUS FUTURE: @dylan_curious

CURIOUS PODCAST: @dontsweatitpod

CURIOUS FRIENDS: @vegasfriends

00:00 - Artificial General Intelligence, 50% Complete: Fact or Fiction?
00:42 - Did Google’s Invention of the Transformers Take Us 20% of the Way to AGI?
01:23 - 30% to AGI in 2020? Unraveling the AI Progress Debate
02:24 - AI Designing AI: The 31% Milestone Towards AGI in 2021
03:14 - DeepMind’s Gato Model: The Super Multitasker Pushing AGI to 39%?
03:48 - We Are 39% Of The Way To AGI: Unraveling Antropic’s Human Reinforcement Method
04:55 - Microsoft’s ChatGPT Pushed Us 41% Closer To AGI
05:49 - AGI at 42%: After Google’s PaLM-E 562B’s Shows Adaptive Abilities
07:19 - AGI at 48%: What Does this Mean for AI’s Future?
08:43 - Humans at 94%, AI at 90.9% | AGI Progress Is Moving Fast
09:37 - 2 Milestone To Look For Before AGI Arrives On Earth

SOURCES:
@MachineLearningStreetTalk
@DrAlanDThompson
@msftautonomoussys
@Structon
@AgilityRobotics

WATCH THE FULL VIDEO ⤵

#ai #artificialintelligence #tech #agi #gpt3 #gpt4 #deepmind #chatgpt #robotics #google #microsoft #machinelearning #autonomousai #palm-e562b #gato #transformers #futuretech #innovation #aitechnology #aitimeline #curiousfuture #alanthompson #artificialgeneralintelligence #aidesign #aipredictions #aiadvancements #aiupdates

[Learna_Hydralis]

I think by exponential nature of development the other 50% would be faster than the first.

[srhatfield]

6:48 It is only a matter of time before the robot decides that the person putting the chips back in the drawer is the problem and needs to be eliminated...

[matthewforsee]

I was so surprised when after I watched your video you didn’t already have a million subscribers!! Very top notch production and writing paired with a natural comedic timing. I am positive you will be at that subscriber count soon! I hope you don’t think this was written by an artificial intelligence.

[travisporco]

why is that identity matrix freaking out

[megret]

A favourite Rick & Morty clip is the butter-passing robot realises how lowly it is

[edh]

The best thing an AGI could do for humanity is to prevent us from killing each other, not by force, but by disrupting supply chains, communications, and financial transactions that enable the military machines throughout the world.

[MickeyMouse-puxt]

Oh my god your channel is so underrated man. And the content is top notch I really liked it

[joannot]

It's fine to use AI to write your script, but it's cringe when you don't check the output and are nothing but a parrot repeating whatever it tell you.
Only an AI would confidently write that anthropic is a branch of open AI 3:59

[bobtarmac]

I believe we are on borrowed time. An Ai tantrum will end you, your family, all of us. Ai jobloss, Ai as weapons, the list goes on. Can we please find a way to cease Ai / GPT? Or begin pausing Ai before it’s too late?

[caseyarmstrong]

Came back to this. 13K views, kudos! Would love to see summary vids from you for long popular AI YouTuber vids.

[_--_]

The Motherland thanks you for your good service as an informer, comrade!

[vallab]

This AGI progress gives us hope in the future for humanity

[lifes_magic_moments]

AGI and ASI saves humanity. we get AGI in 2 years. ASI achieved by 2029. Human illness and Genetic flaws and longevity resolved by 2030. We all will live to 300 and will like and feel like we did at 21
Cheers and see you guys there. Enjoy the ride

[Joe_Brig]

CC heard you say "Open Eye". I think that maybe a more fitting name for OpenAI.

[gcmisc.collection]

My main point is, while developing software and use. The process of an evaluation can be applied to none organic. With the cleaver monkeys, adding to, and manipulating code [AI's DNA] . Plus the software to (learn) compute, ie new human languages. The whole separate system speaking in the same binary. I appreciate different platforms, devices, gateways security.
It just seems an obvious progress for the binary to speak to binary (not sentient ). Small example Pi can recognize a humans text digital fingerprint on any device or platform.. VR headsets software can recognize any human s unique body movements..
The Machine don't now what that means it just links.. VERY SOON the other binary links will form. As in Evolution
.. Trouble is no one, not even the software knows what is Evolving.

[carlhopkinson]

This video is informative and beautiful. Very very well done !!! Subscribed.

[russyehl]

That sounds like Utopia where is the yin with the yang

[Kneephry]

My test for AGI rather than the coffee: a robot that can load a moving van with all of someone's stuff and furniture without hurting anyone or breaking anything.

[j.d.]

We have made it half way through within a few years, that means the second half might not take a couple of months.

[foreropa]

The one with the chips will be one of the first to be eliminated as the ones who made to trip the Boston Dynamics robots