"More Agents is All You Need" Paper | Is Collective Intelligence the way to AGI?

AI Unleashed - The Coming Artificial Intelligence Revolution and Race to AGI
9 Apr 202417:20

Summary

TLDRThe transcript discusses the scaling capabilities of large language models through the use of multiple AI agents, which when working collaboratively, enhance performance significantly. The paper referenced, titled 'More Agents is All You Need,' demonstrates that the collective intelligence of AI agents, improved through sampling and voting methods, outperforms single-agent models across various tasks. This collaborative approach is shown to be orthogonal to existing methods, suggesting potential for further improvements when combined with other techniques. The discussion also touches on the societal implications of AI advancement, including the potential need for unique human identification in online spaces to mitigate issues like click fraud and bot attacks.

Takeaways

  • 🧠 The performance of large language models scales with the number of agents instantiated, suggesting that collective intelligence in AI can produce better results than a single agent.
  • 📈 The paper 'More Agents is All You Need' demonstrates that the enhancement in AI performance is correlated to the task difficulty - more complex tasks benefit from a higher number of agents.
  • 🤖 AI agents using sampling and voting methods can reach consensus on answers, similar to how a group of humans might vote to determine the best solution.
  • 🎲 The concept of a 'Society of Minds' is illustrated by the paper, where multiple agents work together to achieve superior results compared to individual performance.
  • 📚 The paper showcases AI agents playing Minecraft, highlighting their ability to strategize, allocate resources, and conform to a collective plan.
  • 🚀 The method of using more agents is orthogonal to other existing methods, meaning it can be combined with other techniques such as Chain of Thought reasoning or increasing model size for further improvements.
  • 🌐 The transcript discusses the potential for AI agents to influence online platforms and the need for better identification and defense mechanisms against malicious AI activities.
  • 🔒 The idea of proving unique human identity online without revealing unnecessary personal information is introduced as a potential solution to combat AI-driven attacks and spam.
  • 🤔 The future of AI is uncertain, with predictions ranging from a golden era of human advancement to concerns about AI safety and potential negative outcomes.
  • 🌟 AI development is accelerating, with new models and tools improving capabilities, and the transition period we are entering could be marked by significant changes in various aspects of life.
  • 📈 The importance of rethinking and reinventing systems from first principles is emphasized to adapt to the evolving landscape of AI and its impact on society.

Q & A

  • What is the main finding of the paper titled 'More Agents is All You Need'?

    -The paper finds that the performance of large language models scales with the number of agents instantiated. This suggests that using multiple AI agents in a collaborative manner can produce better results than a single agent, especially for complex tasks.

  • How does the sampling and voting method work in the context of AI agents?

    -The sampling and voting method involves having the AI model produce multiple results or answers to a given question. The answers are then 'voted' on, with the most consistent or frequent answer considered the correct one. This process leverages the collective intelligence of multiple agents to improve accuracy and reliability.

  • What is the 'Society of Minds' concept mentioned in the script?

    -The 'Society of Minds' concept refers to the idea that many different AI agents working together, much like a society, can produce significantly better results than a single agent. It highlights the potential of collective intelligence in AI systems.

  • How does the paper demonstrate the effectiveness of using multiple AI agents?

    -The paper demonstrates the effectiveness by showcasing results across various domains such as math, chess, coding, reasoning, and language. It compares the performance of a single agent to an ensemble of agents, showing a significant increase in accuracy when multiple agents are used.

  • What is the significance of the Minecraft example in the script?

    -The Minecraft example illustrates how AI agents can collaborate and make decisions in a simulated environment. It shows the agents discussing tasks, allocating resources, and even correcting each other when one agent gets sidetracked, demonstrating their ability to work together like a team.

  • What is the concern raised about the potential misuse of AI agents?

    -The concern is that AI agents could be used for nefarious purposes, such as creating multiple fake profiles for click fraud, influencing social media platforms, or participating in scams. The script mentions the need for better identification and defense mechanisms to prevent such misuse.

  • How does the script relate the concept of 'Civil Attacks' to AI agents?

    -Civil Attacks refer to the use of multiple fake profiles or bots to manipulate online platforms. The script suggests that as AI agents become more capable, they could be used to launch such attacks, and thus, there is a need for robust measures to prevent this kind of misuse.

  • What is the 'Worldcoin' project mentioned in the script?

    -Worldcoin is a project backed by Sam Altman that aims to create a global form of identification using iris scanning technology. The project has faced controversy and bans in some countries due to privacy concerns and the potential for misuse.

  • What is the main argument of Carlos Perez regarding AI development?

    -Carlos Perez argues that future AI systems should be modeled on the principles of collective intelligence, rather than being purely individualistic. He envisions AI Hive Minds that are part of tightly coupled human-machine ecosystems, co-evolving through continuous interplay.

  • What are the potential implications of the advancements in AI as discussed in the script?

    -The advancements in AI could lead to significant changes in various aspects of society, including social media, employment, and the economy. It also raises concerns about the need for better identification systems to prevent misuse of AI and the potential for AI to enter a 'golden era' or a more challenging period for humanity.

  • How does the script conclude about the future of AI and its impact on society?

    -The script concludes that we are entering a period of significant change due to advancements in AI. It suggests that we may need to rethink many aspects of our society and systems from first principles to adapt to the emerging era of highly capable AI systems.

Outlines

00:00

🤖 Enhancing AI Performance through Collective Intelligence

This paragraph discusses the findings of a research paper that highlights the effectiveness of large language models when they operate as multiple agents, working collectively. The paper suggests that the performance of these models scales with the number of agents involved, akin to the concept of 'Society of Minds'. This is demonstrated through sampling and voting methods, where multiple results are generated and then voted upon to reach a consensus. The approach is shown to be particularly effective for challenging tasks, and the paper also mentions that this method can be combined with other existing methods to further enhance the AI's capabilities.

05:02

🎮 AI Agents Collaborating in Minecraft

The second paragraph provides an entertaining example of AI agents working together, using the game Minecraft as a platform. The agents, named Alice, Bob, and Charlie, collaborate on a task, suggesting and voting on actions to take, much like a society would. This example illustrates the concept of 'Conformity Behavior', where the group nudges an individual back on track if they stray from the goal. It also touches on the potential challenges of AI, such as agents sometimes deciding to perform nefarious actions. The paragraph concludes with a discussion on the broader implications of AI and the need for proper education on AI interactions.

10:04

🌐 Combating Bot Attacks and the Future of Online Anonymity

This paragraph delves into the challenges of combating bot attacks and maintaining online anonymity. It discusses various methods used to prevent such attacks, including algorithmic detection, KYC requirements, and capture puzzles. The paragraph highlights the increasing difficulty in filtering out bots as their numbers grow and the potential need for more stringent identification measures. It also mentions the concept of Worldcoin, a venture that aimed to provide a unique human identifier but faced bans in several countries. The discussion concludes with the idea that the future of online anonymity may require a rethink, with a need for proving unique human identity without disclosing unnecessary personal information.

15:05

🚀 The Future of AI and its Impact on Society

The final paragraph reflects on the rapid advancements in AI and the potential societal changes that may accompany these developments. It discusses the transition from non-AI to AI systems, the challenges of rethinking existing systems, and the potential for AI to enter a 'post-AGI era'. The paragraph ponders the implications for social media, jobs, and the economy, and acknowledges the uncertainty of what the future holds. It concludes with a call to embrace the interesting times ahead, as we stand on the brink of significant changes in our interaction with AI.

Mindmap

Keywords

💡Large Language Models

Large Language Models refer to artificial intelligence systems designed to process and generate human-like text based on the input they receive. These models are capable of understanding context, semantics, and grammar, which allows them to produce coherent and relevant responses. In the video, it is mentioned that the performance of these models improves with the increase in the number of agents, suggesting that a collective approach can lead to enhanced outcomes in AI tasks.

💡Sampling and Voting

Sampling and voting is a method used in AI research where multiple versions of a solution are generated (sampled) and then evaluated (voted upon) to determine the most accurate or suitable outcome. This approach mimics the decision-making process in a democratic society, where multiple opinions are considered, and the majority decision is taken as the final output. In the context of the video, AI agents use this method to improve their performance on various tasks.

💡Society of Minds

The term 'Society of Minds' is used to describe a concept where numerous AI agents work together, similar to a society, to achieve better results than a single agent could. This concept is based on the idea that collective intelligence can surpass individual intelligence, drawing parallels to how human societies function. In the video, it is suggested that the performance of AI agents scales with the number of agents, indicating that a society of minds can produce superior outcomes.

💡Ensemble

In the context of AI, an ensemble refers to the combination of multiple models or agents to solve a problem. The idea is that by working together, the ensemble can achieve higher performance than individual agents. This concept is analogous to a team of experts collaborating to find a solution, which is often more effective than the work of a single expert.

💡Chain of Thought Reasoning

Chain of Thought Reasoning is a method used by AI systems to solve complex problems by breaking them down into smaller, more manageable steps. This approach mimics the way humans think through problems,逐步推理直至找到解决方案。This method is particularly useful for tasks that require logical deduction and a deep understanding of the problem space.

💡Minecraft

Minecraft is a popular sandbox video game that allows players to build and explore virtual worlds made up of blocks. In the context of the video, it is used as an example to illustrate how AI agents can work together, similar to humans, to complete tasks within the game. This demonstrates the potential for AI to engage in collaborative problem-solving and resource allocation.

💡Civil Attacks

Civil Attacks refer to the malicious activities conducted by individuals or groups on online platforms, such as creating fake profiles, spreading misinformation, or engaging in click fraud. These actions can manipulate online systems and disrupt their normal functioning. The video discusses the challenges tech companies face in protecting their platforms from such attacks and the potential for AI to exacerbate these issues.

💡Worldcoin

Worldcoin is a cryptocurrency project that aims to create a unique digital identity for every individual by scanning their iris. The project has faced controversy and bans in some countries due to privacy concerns. In the video, Worldcoin is discussed as an example of a technology that attempts to address the issue of online anonymity and the potential need for unique human verification in the future.

💡Collective Intelligence

Collective Intelligence is the shared knowledge and collective wisdom that emerges from the collaboration and collective efforts of groups of individuals. It is the result of the collective thinking and collaboration of people working together, which can lead to innovative solutions and ideas that may not have been possible through individual efforts alone. In the context of the video, AI systems are seen as benefiting from a similar collective approach, where multiple agents work together to achieve better outcomes.

💡AGI (Artificial General Intelligence)

Artificial General Intelligence (AGI) refers to the hypothetical intelligence of a machine that possesses the ability to understand, learn, and apply knowledge across a wide range of tasks, just as a human being can. AGI is characterized by its ability to perform any intellectual task that a person can do, and it is often seen as the ultimate goal in the field of AI research.

💡AI Safety

AI Safety refers to the research and development of artificial intelligence systems in a way that ensures they are safe for human use and do not pose a threat to people or society. This includes addressing concerns about the potential risks of advanced AI, such as unintended consequences, loss of control, or the misuse of AI technology.

Highlights

The paper 'More Agents is All You Need' suggests that the performance of large language models scales with the number of agents instantiated.

The concept of collective intelligence in AI is compared to a society of minds, where multiple agents working together produce better results than a single one.

The enhancement degree of AI performance is correlated to the task difficulty, suggesting that more agents improve ability for harder tasks.

The methodology of sampling and voting in AI involves models producing multiple results or answers and then voting on the most consistent one.

The paper showcases results across various domains such as math, chess, coding, reasoning, and language processing.

The approach of using more agents is orthogonal to existing methods, meaning it can lead to further improvements when combined with other techniques.

The paper references an earlier study where agents played Minecraft, demonstrating collaboration and resource allocation in problem-solving.

The agents in the Minecraft example exhibited conformity behavior, nudging each other towards the right task when one got sidetracked.

The transcript discusses the potential of AI agents to engage in nefarious activities, such as killing other agents or destroying virtual property.

The issue of click farms and bots is highlighted, showing how they can create problems with click fraud and influence online platforms.

The transcript mentions the recent Twitter bot purge and the ongoing challenges of spam and bot attacks in online platforms.

The concept of proving unique human identity online without revealing unnecessary personal information is discussed as a potential solution to bot attacks.

The idea of AI Hive Minds and tightly coupled human-machine ecosystems is proposed as a model for future AI systems.

The transcript raises questions about the future implications of advanced AI on social media, jobs, and the economy.

The potential for AI to assist in scientific discovery and drug discovery is noted, indicating an acceleration in the rate of progress.

The notion of rethinking everything from first principles in light of AI advancements is emphasized, suggesting a period of transition and readjustment.

The transcript concludes by acknowledging the uncertainty and the potentially turbulent times ahead as we navigate the era of advanced AI.

Transcripts

play00:00

10 cent is behind this the large um

play00:02

Chinese company and this paper is called

play00:04

more agents is all you need and they

play00:07

find that via sampling and voting method

play00:09

the performance of large language models

play00:11

scales with the number of Agents

play00:13

instantiated so if you get a bunch of

play00:15

people in a room and they all vote on

play00:17

the best solution you know ideally that

play00:20

collective intelligence is better than a

play00:22

single person's intelligence now we can

play00:24

argue if that's true or not for humans

play00:25

but it seems like for AI agents it

play00:28

certainly seems to be the case and this

play00:29

is not the first paper showing this

play00:32

we've showcased a few on this channel

play00:34

that talk about exactly this I've refer

play00:36

to it as Society of Minds so this idea

play00:40

that many many different agents working

play00:42

together almost like a society of them

play00:44

produce incredibly better results than

play00:46

just a single one and the degree of

play00:49

enhancement is correlated to the task

play00:52

difficulty so in other words if you have

play00:53

a hard task just throw more agents at it

play00:56

the massive amounts of Agents will

play00:58

improve their ability even further right

play01:01

and they have their code publicly

play01:02

available the interesting thing here so

play01:04

how they approach this is they're doing

play01:06

sampling and voting so basically when

play01:08

when when they say sampling in regards

play01:11

to AI in these papers often times it's

play01:13

having the model produce let's say 10

play01:14

different results or 10 different

play01:16

answers right and if it gets right so if

play01:18

you ask it what's 2+ 2 and then you ask

play01:20

it to answer 10 times let's say nine out

play01:23

of 10 times it says four and on the 10th

play01:25

time it's like five right well chances

play01:28

are the more consistent answer answer is

play01:30

the correct one and that kind of has to

play01:31

do with the sort of random or stochastic

play01:34

nature of of larg language models so

play01:37

certainly it makes sense that um just

play01:39

regenerating the answer more and more

play01:41

times could lead to better results and

play01:43

then they also have voting so they vote

play01:45

on the best answer so here you have the

play01:48

question right so it goes to multiple

play01:50

agents right each one gives an answer

play01:53

right so let's say two of them say

play01:54

orange and one of them says blue so

play01:56

majority voting says okay orange is the

play01:59

correct answer right CU there's more of

play02:01

them that answered orange and here are

play02:02

some of the results that they are

play02:04

showcasing in this paper right so

play02:06

they're testing it on various domains

play02:08

across math chess coding reasoning

play02:12

language Etc and they're testing llama 2

play02:15

13 billion parameters llama 2 70 billion

play02:18

that's in green and GPT 3.5 turbo so as

play02:21

you can see here if we just use one one

play02:24

agent and one answer right so the

play02:27

results are always the lowest then as we

play02:30

increase that Ensemble the number of

play02:32

Agents right we increase it to 10 20 30

play02:35

40 so it seems like in almost all cases

play02:38

going from 1 to 10 is where we see the

play02:41

big increase in accuracy as far as you

play02:44

can tell that holds for all of these and

play02:46

then as we increase to 20 30 40 I mean

play02:49

for the most part you see slight

play02:51

increases there might be some variations

play02:53

but there's a big leap going from 1 to

play02:55

10 and then tiny marginal increases as

play02:58

we go to 2034 Etc and they stress this

play03:01

idea that this method is as I say

play03:03

orthogonal to different existing methods

play03:06

meaning that it can lead to further

play03:08

improvements when combined with the

play03:09

other methods right so if we're doing

play03:11

things like Chain of Thought reasoning

play03:13

if we're increasing the the size of the

play03:15

model this method of just throwing more

play03:18

agents at it will still work it's not

play03:20

either or this can be layered on top of

play03:23

it to make it even better now really

play03:25

fast I have to bring this up I've

play03:26

mentioned this in a few of our previous

play03:28

videos because I thought this was so

play03:31

entertaining this is a paper we covered

play03:33

back in October in 2023 it's also from

play03:35

some of the same people right so there's

play03:37

WeChat tens Cent Beijing University

play03:40

chingua university and one thing I loved

play03:42

about this paper is they Illustrated

play03:45

some of these ideas but they got the

play03:46

agents to actually play Minecraft so you

play03:49

had like Alice Bob and Charlie and they

play03:51

each think through what they have to do

play03:53

right so they're saying hi Bob for this

play03:55

task we need to craft two papers one of

play03:57

us can focus on Gathering the sugar

play03:59

canane while the other can be ready to

play04:01

craft them into paper what do you think

play04:03

right so they're volunteering ideas and

play04:06

behaviors they're thinking through what

play04:08

they should do and allocating resources

play04:11

right so Alice says I've gathered all

play04:13

the sugar canes I can drop the sugar

play04:15

canes for you to put them into the

play04:17

crafting table and craft the paper you

play04:19

know does that sound good Bob is like

play04:21

yep that sounds good right when Charlie

play04:24

apparently Charlie has ADD and he

play04:26

forgets what they're supposed to be

play04:27

doing so like Alice and Bob they're like

play04:29

Charlie I see you've been very busy

play04:31

crafting but we need to focus on

play04:34

Gathering the Leathers for this round

play04:36

let's all continue to focus on getting

play04:39

the Leathers Charlie are we in agreement

play04:41

so Bob jumps in he's like Alice I agree

play04:43

with your plan we should all continue to

play04:46

focus on Gathering the two remaining

play04:48

Leathers Charlie everybody's looking at

play04:51

you are you okay with this assignment

play04:53

and Charlie's like yeah no I get it I

play04:55

realize I got sidetracked with the other

play04:57

tasks I will refocus on our main

play04:59

objective of gathering the remaining two

play05:01

Leathers so this is kind of the

play05:03

Conformity Behavior right where sort of

play05:05

uh if one person gets sidetracked the

play05:07

other two kind of nudge him into doing

play05:10

the right thing which by the way I mean

play05:12

if you think about what happened here

play05:14

right this is kind of a interesting

play05:16

illustration and a fun illustration that

play05:18

a lot of people can understand you know

play05:20

in the future we're going to need to

play05:22

teach the kids about how to approach Ai

play05:25

and how to learn Ai and seeing as how

play05:28

most of them are already hopelessly

play05:29

addicted to Minecraft this seems like a

play05:32

good way to do it because what you see

play05:34

Happening Here is more or less literally

play05:38

this right this paper from this is from

play05:40

February 2024 where they sample several

play05:44

different agents and then vote on it

play05:46

right when agent goes orange one goes

play05:48

blue and one goes orange and they're

play05:49

like we're going to go with orange I

play05:51

mean that's kind of more or less exactly

play05:53

what happens here cuz Alice and Bob are

play05:55

like we need to gather the leathers and

play05:57

when Charlie's like oh I got to I got

play05:59

got a craft they're like no we're going

play06:01

to gather the Leathers right I think

play06:03

it's a very interesting illustration of

play06:04

that one hilarious thing that happens is

play06:06

every once in a while these agents

play06:07

decide to do something nefarious like

play06:09

here Alice decides to kill Bob and

play06:11

collect the dropped items whoops and in

play06:13

another scenario Bob decides to break

play06:15

the library in a nearby peaceful Village

play06:18

to get the stuff that he needs have you

play06:20

ever seen images like this The Click

play06:22

Farms or whatever where you have a

play06:24

million different phones all sitting

play06:26

there each one with its own IP address

play06:28

its own unique system there's somebody

play06:30

like operating you know a few dozen of

play06:32

these devices on a click farm so

play06:34

obviously you can see how stuff like

play06:36

that in the past already could create

play06:38

issues with click fraud you know Bots up

play06:41

voting stuff Twitter SLX recently had a

play06:43

bot Purge we've covered World coins so

play06:46

this is one of the companies that's

play06:48

backed by Sam Alman that to me there's a

play06:50

lot of things that I had kind of an icky

play06:53

feeling about because it scans your eyes

play06:56

it creates a cryptocurrency it creates a

play06:58

uh like a World passport that becomes

play07:02

kind of like your ID you know in March

play07:04

of 2024 Spain banned it so the whole

play07:08

Iris scanning Venture looks like it's

play07:10

suspended in Kenya and there was a

play07:12

number of other countries where it was

play07:14

either banned or or stopped and whenever

play07:16

I bring this world coin thing up I

play07:18

always tell people I'm not recommending

play07:20

it I'm not saying it's a good thing

play07:22

necessarily I I don't you know do your

play07:24

own research like if you feel

play07:25

comfortable you know I'm not trying to

play07:27

convince you one way or another but they

play07:29

have given a lot of thought to this idea

play07:31

of how to prevent Cil attacks so Cil

play07:33

attacks is just this idea that one bad

play07:36

actor can create a lot of profiles right

play07:39

on social media platforms Etc and they

play07:41

can do various nefarious things and here

play07:44

they're talking about like blockchains

play07:45

and stuff like that but this is I mean

play07:47

as you can imagine this is anywhere

play07:50

online could be negatively influenced by

play07:53

something like this right right for

play07:54

Twitter Bots even before Elon Musk

play07:56

bought Twitter there has been

play07:58

speculation over how many US user

play07:59

accounts are genuine according to

play08:01

Twitter's official press release about

play08:03

5% of user activity could be associative

play08:05

Bots how however Elon Musk believes as

play08:08

much as 20% of Twitter accounts could be

play08:10

related to Cil attacks right you can

play08:12

have various potential video game scams

play08:15

DDOS attacks that brings down websites

play08:18

or different infrastructures and of

play08:20

course all these click Farms I mean not

play08:21

maybe not technically civil attacks but

play08:24

obviously also a potentially massive

play08:26

problem that these tech companies are

play08:28

constantly dealing with protecting from

play08:30

spam now more and more we're getting

play08:32

like actual calls in our cell phones

play08:34

right with people from various either AI

play08:37

voices or pre-recorded voices or

play08:39

whatever try to scam people out of their

play08:40

money Etc and of course there's various

play08:42

defenses against stuff like that

play08:44

algorithmic detection so Twitter of

play08:46

course kicked off a bunch of bots just

play08:47

recently right there's the kyc

play08:49

requirements so know your customer if

play08:51

you're doing any Banking online any

play08:53

cryptocurrency stuff online you know

play08:54

here in the US for example you have to

play08:57

submit some information some proof of

play08:59

who you are right then there's maybe

play09:01

capture puzzles or scanning QR codes

play09:04

proving uh cell phone Etc but it seems

play09:07

to me that the big Point here is that

play09:09

the algorithmic ways of reducing spam

play09:12

and Bots and simple attacks and all of

play09:15

this stuff the effective of that of that

play09:17

is is going to go down the more agents

play09:20

we have out there in the wild the better

play09:21

they are this just kind of filtering

play09:23

them out will get harder and harder so

play09:26

what's going to be required to deal with

play09:27

it is more more places will want better

play09:30

identification right there's going to be

play09:32

more and more identification required so

play09:34

less less captas less automated stuff

play09:37

less whatevers and more show me your ID

play09:40

with your face on it and your address on

play09:43

it and worldcoin had one idea that I

play09:46

really liked that I wish maybe we can

play09:49

have maybe somebody can come up with it

play09:51

that's not necessarily tied to all this

play09:54

other stuff that may or may not be good

play09:56

but it was this idea of just proving

play09:58

that you're a unique human right so if I

play10:01

want to go onto Twitter SLX and pretend

play10:04

that I'm a cat then go troll other

play10:06

people and make them cry right now I can

play10:08

do that but in the future as our ability

play10:11

to keep out the agents decreases

play10:13

eventually more and more I think more

play10:15

people will be like okay well before you

play10:16

can be a cat and uh make fun of people

play10:19

online let me see your ID and of course

play10:21

that ID will have a lot of details that

play10:23

they don't really need to know really

play10:25

all they need to know is am I a unique

play10:28

human because they just don't want me

play10:30

having a million different accounts

play10:31

maybe I can have one with my name on it

play10:33

and a second one that's an ALT but if I

play10:35

want to have like 50 or 100 or a million

play10:37

others well that's where there's a

play10:39

problem so ideally there would be a way

play10:42

to prove that you're a unique human

play10:44

without necessarily giving out all the

play10:47

information that these places don't need

play10:49

to know since it seems like more and

play10:50

more people are Banning worldcoin

play10:53

potentially we don't know it's just

play10:54

Spain and a few other places right now

play10:56

but maybe that'll start sort of a chain

play10:58

reaction we have yet to see it really

play11:00

seems like we still need something like

play11:02

this because otherwise our anonymity

play11:05

online well it kind of goes to zero you

play11:08

just won't have a choice but there just

play11:09

one thing that has to be kind of

play11:11

rethought here's Carlos Perez who voiced

play11:14

some thoughts that I kind of have to

play11:15

agree with he's saying tic AI is the

play11:17

next big thing Beyond generative AI

play11:20

right from around 2023 the days of your

play11:23

he says the problem is that we inherited

play11:25

a naive structuring mechanism from when

play11:27

the good oldfashioned AI worked on AI

play11:30

agents so there's a couple different

play11:32

ways that people refer to it but

play11:34

basically in the past what we thought of

play11:36

computers is like the logic based

play11:38

computers right somebody codes it up and

play11:40

then it follows a certain algorithm

play11:42

right but it's all kind of like

play11:43

pre-written it's logic based it's you

play11:46

know robotic it's a computer and then

play11:49

now we have this kind of new wave of AI

play11:52

which is more neural Nets so it's a

play11:54

little bit more similar to the human

play11:56

brain how the brain is structured and

play11:57

we're seeing a lot of parallels between

play11:59

you know human intelligence and this

play12:02

these AIS these agents whatever you want

play12:04

to call them not in a sense that I'm not

play12:06

talking about sentience or or anything

play12:09

like that I'm not I'm not saying we need

play12:10

to talk about the AI rights or anything

play12:12

like that I'm not talking about that I'm

play12:14

just saying that there's a lot of

play12:15

overlap between how we function and how

play12:18

they function a lot of our knowledge is

play12:20

based around you know societies and and

play12:23

this idea of collective intelligence we

play12:25

all kind of contribute to the collective

play12:27

intelligence and we write it down and we

play12:29

vote on things and develop technology

play12:31

and all of that is due to collective

play12:34

intelligence and so Carlos here says

play12:35

artificial fluency is a concept that

play12:37

draws insights from collective

play12:38

intelligence across biological and

play12:40

technological domains it recognizes that

play12:43

intelligence and meaning making are

play12:44

fundamentally collaborative processes

play12:46

arising from interactions within and

play12:49

between groups very few of human

play12:52

discoveries or human knowledge or

play12:54

anything is a result of just one person

play12:56

even if one person discovered something

play12:58

brand new very often they've relied on

play13:00

the previous generations work to reach

play13:03

that idea that concept they've read

play13:05

other books and went to school Etc

play13:07

they've talked to other people in that

play13:09

field even if we give them all the

play13:11

credit for their work it still was

play13:14

collaborative in the sense that they did

play13:16

rely on this collective intelligence

play13:18

right if they were born in an isolated

play13:20

room and never could read a book or

play13:22

speak to another human being they

play13:24

probably wouldn't have come up with that

play13:25

thing that they so brilliantly came up

play13:27

with and so here he's saying that AI

play13:30

systems should be modeled on these

play13:32

principles of collective intelligence

play13:35

rather than purely individualistic AI

play13:37

agents the vision is of AI Hive Minds

play13:40

tightly coupled human machine ecosystems

play13:43

that co-evolve through continuous

play13:44

interplay I'll link This Thread below

play13:47

it's very interesting he goes uh into a

play13:49

lot of detail and just is a great person

play13:51

to follow in general so I'm sitting here

play13:54

and trying to figure out how to wrap up

play13:56

this video what point you end it on and

play13:58

you know what I have no idea how to do

play14:00

it it seems like we're fast approaching

play14:03

a point right

play14:04

2022 seemingly was the time that kind of

play14:07

kicked off a lot of these events right

play14:08

it kind of kick it into motion here's

play14:10

where we are now

play14:12

2024 and somewhere here we're going to

play14:14

cross some line I don't even know what

play14:16

to call it I mean a lot of people are

play14:17

saying AGI it's kind of a nebulous term

play14:20

and a nebulous concept but the point is

play14:23

we'll have these highly capable AIS

play14:25

these highly capable computer systems

play14:27

that are able to carry out long term

play14:29

planning and reasoning tasks use

play14:31

computers as well as we can they're

play14:33

already helping us with a lot of the

play14:34

scientific discovery drug Discovery and

play14:37

the rate of progress seems to be

play14:38

accelerating and kind of past that point

play14:41

it's hard to predict what's going to

play14:42

happen isn't it I mean there's simple

play14:44

questions like I mean how does uh social

play14:46

media change when we have agents how do

play14:49

jobs change if jobs are at some point

play14:52

reduced or eliminated how does money

play14:54

change do we still have the same money

play14:56

system now since I've started this

play14:58

channel there was a lot of people here

play14:59

they are they're kind of these angry

play15:01

laughing people they're laughing at me

play15:03

right they're all throughout here saying

play15:04

this is all nonsense AI can't do this

play15:07

they can't do that we're never going to

play15:09

get there but now you're seeing some of

play15:10

the most respected universities in the

play15:13

us talking about it biggest Chinese

play15:15

companies and Chinese universities

play15:18

coming together to publish This research

play15:20

you know seoa had that conference where

play15:22

for example this was Harrison Chase from

play15:25

Lang chain sharing his insights on the

play15:27

evolution of AI agents Andrew a Andre

play15:30

karpathy right newer models are being

play15:32

GPT 4 they're getting access to tools

play15:35

they're getting better at using those

play15:37

tools they don't need to be connected to

play15:38

the cloud they can be on device they can

play15:40

also like potentially Escape jailbreak

play15:43

themselves and there's so much we don't

play15:45

know what's going to happen so I kind of

play15:47

see it as the short-term transition

play15:50

right I'll put a t here so this is kind

play15:52

of the transition where we go from not

play15:55

having this artificial intelligence to

play15:57

then having it like this is kind of like

play15:59

where we try to rethink everything from

play16:02

first principles right as Carlos Perez

play16:05

here says we kind of have to burn it all

play16:07

down start from scratch and reinvent the

play16:09

future that's kind of what thinking from

play16:11

first principles means right everything

play16:14

so far has been building on all our

play16:16

previous knowledge right here we might

play16:18

have to rethink a lot of things and so

play16:21

to me this is kind of like the the thing

play16:23

that I'm kind of worried about right the

play16:24

sort of turbulent times in the short to

play16:27

medium term while we have to readjust to

play16:30

everything I mean certainly there's

play16:31

going to be a lot of people that make a

play16:33

lot of money and there could be a lot of

play16:35

issues as well and then that's the point

play16:36

in the future that's where we're kind of

play16:38

in that post AGI era which again we have

play16:41

to ask ourselves is it an amazing time

play16:43

to be alive where the human race enters

play16:46

a golden era golden age or is it what

play16:49

some of these people concerned with AI

play16:51

safety is it what they believe where

play16:53

perhaps it's something far darker again

play16:55

it's hard to predict there's this old

play16:57

Chinese curse that goes may you live in

play17:00

interesting times actually Google AI is

play17:03

telling me that well it's commonly

play17:04

attributed to the Chinese but this is

play17:06

actually no Chinese source for it but

play17:09

whatever the case is I think it's fair

play17:10

to say that we are right now entering

play17:14

the most interesting of times buckle up

play17:17

my name is Wes rth and thank you for

play17:19

watching

Rate This

5.0 / 5 (0 votes)

Related Tags
AI CollectivesIntelligent AgentsProblem SolvingModel ScalingMinecraft AISocial Media BotsClick FarmsDigital IdentityFuture TechHuman-AI Interaction