Recent Comments by L0cky subscribe to this feed

Largest ever 3D map of the universe

L0cky says...

Good explanation.

A way to demonstrate it is to inflate a balloon a little, draw dots on it at intervals, then inflate the balloon.

From any dot you look at, all other dots are moving away from it.

Dog Loves Jumping In Leaves

Multi-Agent Hide and Seek

L0cky says...

This isn't really true though and greatly understates how amazing this demo, and current AI actually is.

Saying the agents are obeying a set of human defined rules / freedoms / constraints and objective functions would lead one to imagine something more like video game AI.

Typically video game AI works on a set of weighted decisions and actions, where the weights, decisions and actions are defined by the developer; a more complex variation of:

if my health is low, move towards the health pack,
otherwise, move towards the opponent

In this demo, no such rules exist. It's not given any weights (health), rules (if health is low), nor any instructions (move towards health pack). I guess you could apply neural networks to traditional game AI to determine the weights for decision making (which are typically hard coded by the developer); but that would be far less interesting than what's actually happening here.

Instead, the agent is given a set of inputs, a set of available outputs, and a goal.

4 Inputs:
- Position of the agent itself
- Position and type (other agent, box, ramp) of objects within a limited forward facing conical view
- Position (but not type) of objects within a small radius around the agent
- Reward: Whether they are doing a good job or not

Note the agent is given no information about each type of object, or what they mean, or how they behave. You may as well call them A, B, C rather than agent, box, ramp.

3 Outputs:
- Move
- Grab
- Lock

Again, the agent knows nothing about what these mean, only that they can enable and disable each at any time. A good analogy is someone giving you a game controller for a game you've never played. The controller has a stick and two buttons and you figure out what they do by using them. It'd be accurate to call the outputs: stick, A, B rather than move, grab, lock.

Goal:
- Do a good job.

The goal is simply for the reward input to be maximised. A good analogy is saying 'good girl' or giving a treat to a dog that you are training when they do the right thing. It's up to the dog to figure out what it is that they're doing that's good.

The reward is entirely separate from the agent, and agent behaviour can be completely changed just by changing when the reward is given. The demo is about hide and seek, where the agents are rewarded for not being seen / seeing their opponent (and not leaving the play area). The agents also succeeded at other games, where the only difference to the agent was when the reward was given.

It isn't really different from physically building the same play space, dropping some rats in it, and rewarding them with cheese when they are hidden from their opponents - except rats are unlikely to figure out how to maximise their reward in such a 'complex' game.

Given this description of how the AI actually works, the fact they came up with complex strategies like blocking doors, ramp surfing, taking the ramp to stop their opponents from ramp surfing, and just the general cooperation with other agents, without any code describing any of those things - is pretty amazing.

You can find out more about how the agents were trained, and other exercises they performed here:

https://openai.com/blog/emergent-tool-use/

bremnet said:

Another entrant in the incredibly long line of adaptation / adaptive learning / intelligent systems / artificial intelligence demonstrations that aren't. The agents act based on a set of rules / freedoms/constraints prescribed by a human. The agents "learn" based on the objective functions defined by the human. With enough iterations (how many times did the narrator say "millions" in the video) . Sure, it is a good demonstration of how adaptive learning works, but the hype-fog is getting a big thick and sickening folks. This is a very complex optimization problem being solved with impressive and current technologies, but it is certainly not behavioural intelligence.

This is why women love him

L0cky says...

Installation of the Peg Valve Transference Unit requires a signed TN.17 (3 copies); a filled and signed FL.54 (additionally signed by 3 witnesses); majority approval from the Board of Optimal Manufacturing Process; majority approval from the Board of Human Affairs; and a sworn affidavit from the CEO (or appointed delegate if the CEO is on leave) that the proposed change will neither impact the human rights of the human resources, nor negatively impact the economy of the state.

The above solution only required a signed Refrain from Sexual Misconduct and a basket.

AeroMechanical said:

Uh, and if, as appears to be the case, *every* widget goes in the basket, why is there a person doing that?

The difference between a large beer and a small beer

L0cky says...

But were they actually sold as 'big beer' and 'small beer' or simply different brands in different glasses?

I've found people are generally bad at estimating volumes of glasses.

Pizza dance

Spain-1 US Navy-0

Age of Sail

Undeniable Miracle

"Do You Trust This Computer?" Documentary

exurb1a - You (Probably) Don't Exist

L0cky says...

There is a generally held belief that consciousness is a mystery of science or a miracle of faith; that consciousness was attained instantly (or granted by god), and that one has either attained self awareness or has not.

I don't believe any of that. I believe like all things in biology, consciousness evolved to maximise a benefit, and occurred gradually, without any magic or mystery. The closest exurb1a gets to that is when he says at 6:28:

"Maybe evolution accidentally made some higher mammals on Earth self-aware because it's better for problem solving or something"

We need to know what other people are thinking and this is the problem that consciousness solves. If a neighbouring tribe enters your territory then predicting whether they come to trade, mate, steal or attack is beneficial to survival.

Initially this may be done through simulation - imagining the future based on past experience. A flood approaching your cave is bad news. Being surrounded by lions is not good. Surrounding a lone bison is dinner. Being charged by a screaming tribe is an upcoming fight.

We could only simulate another person's actions, but we had no experience that allows us to simulate another person's thoughts. You may predict that giving your hungry neighbour a meal may suppress their urge to raid your supplies but you still can't simply open their head and see what they are thinking.

Then for the benefit of cooperation and coordination, we started to talk, and everything changed.

Communication not only allows us to speak our mind, but allows us to model the minds of others. We can gain an understanding of another person's motivations long before they act upon them. The need to simulate another person's thoughts becomes more nuanced and complex. Do they want to trade, or do they want to cheat?

Yet still we cannot look into the minds of others and verify our models of them. If we had access to an actual working brain we could gradually strengthen that model with reference to how an actual brain works, and we happen to have access to such a brain, our own!

If we monitored ourselves then we could validate a general model of thought against real urges, real experiences, real problem solving and real motivations. Once we apply our own selves to a model of thought we become much better at modelling the thoughts of others.

And what better way to render that model than with speech itself? To use all of our existing cognitive skills and simply simulate others sharing their thoughts with us.

At 3:15 exurb1a referenced a famous experiment that showed that we make decisions before we become aware of them. This lends evidence to suppose that our consciousness is not the driver of our thoughts, but a monitor - an interpretation of our subconscious that feeds our model of how people think.

Not everybody is the same. We all have different temperaments. Some of us are less predictable than others, and we tend to avoid such people. Some are more amenable to co-operation, others are stubborn. To understand the temperament of one we must compare them to another. If we are to compare the model of another's mind to our own, and we simulate their mind as speech, then we must also simulate our own mind as speech. Then not only are we conscious, we are self-aware.

Add in a feedback loop of social norms, etiquette, acceptable behaviour, expected behaviour, cooperation and co-dependence, game theory and sustainable societies and this conscious model eventually becomes a lot more nuanced than it first started - allowing for abstract concepts such as empathy, shame, guilt, remorse, resentment, contempt, kinship, friendship, nurture, pride, and love.

Consciousness is magical, but not magic.

exurb1a - You (Probably) Don't Exist

IMPORTANT! YOU ARE BEING MANIPULATED!!!

L0cky says...

Krishnan proposed "They think they're manipulating me, but I know what I'm doing".

To gain a better understanding of how this is possible, watch this great video about AI that ant posted:

https://videosift.com/video/Do-You-Trust-This-Computer-Documentary

Saying that you can handle the attempts to manipulate you would be like the world's chess, poker and go champions saying they can handle their AI challengers.

In reality, AI doesn't just beat them at their games, it now completely destroys them with ease. Statistically affecting your worldview is a much easier task.

Facebook Employees Explain Struggling To Care ...

L0cky says...

Yup, but they're going for a slightly different angle since the satire they're usually known for has now become indistinguishable from reality.

They could literaly switch to just straight covering the news and nobody would notice.

Good comment from YT:

Plot twist: Onion is no longer a fake news channel

"Do You Trust This Computer?" Documentary



Send this Article to a Friend



Separate multiple emails with a comma (,); limit 5 recipients






Your email has been sent successfully!

Manage this Video in Your Playlists

Beggar's Canyon