Jump to content

Why is MIT's Artificial General Intelligence course so weakly known (agi.mit.edu)?


thoughtfuhk

Why is MIT's Artificial General Intelligence course so weakly known (agi.mit.edu)?  

2 members have voted

  1. 1. I find it disappointing that a course concerning perhaps mankind's last invention is hardly known outside of the machine learning community. Why is MIT's Artificial General Intelligence course so weakly known??

    • People underestimate the importance of AGI
    • People think AGI is not important at all
      0
    • This topic doesn't matter to me
      0


Recommended Posts

I find it disappointing that a course concerning perhaps mankind's last invention is hardly known outside of the machine learning community.  Why is MIT's Artificial General Intelligence course so weakly known??

Edit: I just noticed another thread of mine was recently closed. Please refrain from discussing that thread here.

Edited by thoughtfuhk
Link to comment
Share on other sites

9 minutes ago, studiot said:

You flatter yourself.

 

Perhaps you should look beyond your own discipline?

Think about this carefully:

  1. Narrow Ai/deep artificial neural network powered models can now do individual cognitive tasks. (There's Ai for disease diagnosis that does better than human doctors, Ai for etc etc)
  2. General Ai, called AGI, will likely cover the entire landscape of human cognitive ability, when AGI eventually arises. This means it will aid in doing problems involving thinking, which is all problems!
  3. Why wouldn't a model that can do all human cognitive tasks better, not classify as mankind's last invention? Don't you recognize that AGI concerns all disciplines?
Link to comment
Share on other sites

My neighbour has now been on a gene therapy regime to replace the failing chemotherapy regime she was on.


This was only licenced last November and my neighbour has shown a dramatic recovery since.

 

I doubt that an AI doctor, even combined with all the computer engineers at Microsoft, would be able to replace the human ingenuity that lead to this new form of treatment any time soon.

Link to comment
Share on other sites

1 minute ago, studiot said:

My neighbour has now been on a gene therapy regime to replace the failing chemotherapy regime she was on.


This was only licenced last November and my neighbour has shown a dramatic recovery since.

 

I doubt that an AI doctor, even combined with all the computer engineers at Microsoft, would be able to replace the human ingenuity that lead to this new form of treatment any time soon.

  1. Yes, that AGI may emerge much much later, does not warrant that I was "flattering myself" or being narrow minded about the one particular discipline known as AGI.
  2. At least now you've nicely pointed out that AGI is likely merely a matter of time, than a matter of possibility.

Postscript: Kurzweil predicts human level AGI by 2029.

Postscript-2: It is likely that that new treatment you referred to emerged with the aid of Ai!

Link to comment
Share on other sites

10 minutes ago, thoughtfuhk said:
  1. Yes, that AGI may emerge much much later, does not warrant that I was "flattering myself" or being narrow minded about the one particular discipline known as AGI.
  2. At least now you've nicely pointed out that AGI is likely merely a matter of time, than a matter of possibility.

Postscript: Kurzweil predicts human level AGI by 2029.

Postscript-2: It is likely that that new treatment you referred to emerged with the aid of Ai!

 

You are now an expert on gene therapy?

Here is a quote from your agricultural chemical link, which amply demonstrated you and they do not have the faintest idea what I am talking about.

Quote

“There are two distinct problems in drug discovery: biology and chemistry,” he said. “If you’re working on biology, you’re trying to decide which disease protein is the best one to target.  A lot of AI companies in drug discovery are working on this target identification problem. Once you’ve chosen a target, you can start working on chemistry problems: how to deliver a non-toxic molecule that can hit the chosen disease protein.  Atomwise is focused on these chemistry problems; specifically, Atomwise invented the use of deep neural networks for structure-based drug design.”

 

Edited by studiot
Link to comment
Share on other sites

Just now, Strange said:

So we should refrain from discussing AI on a thread about AI?

Reported.

This thread is concerned with MIT's AGI course, not the purpose of human life. 

Report at your leisure, albeit.

5 minutes ago, studiot said:

 

You are now an expert on gene therapy?

Here is a quote from your agricultural chemical link, which amply demonstrated you and they do not have the faintest idea what I am talking about.

 

Sorry, that was the wrong link.

Correct link: "Artificial Intelligence takes Gene Therapy to the next level".

 

PostscriptI didn't say I was an expert at gene therapy. I simply underlined that Ai may aid in problems involving thinking, which is all problems, including gene issues.

Link to comment
Share on other sites

57 minutes ago, thoughtfuhk said:

I find it disappointing that a course concerning perhaps mankind's last invention

Why do you think this is mankind's last invention?

57 minutes ago, thoughtfuhk said:

is hardly known outside of the machine learning community

That is probably true of most specialised courses.

58 minutes ago, thoughtfuhk said:

Why is MIT's Artificial General Intelligence course so weakly known??

Glad to see you are bringing the same high standards of English to this thread.

10 minutes ago, thoughtfuhk said:

This thread is concerned with MIT's AGI course, not the purpose of human life. 

Oh. I thought developing AI was the purpose of human life?

Link to comment
Share on other sites

14 minutes ago, thoughtfuhk said:

This thread is concerned with MIT's AGI course, not the purpose of human life. 

Report at your leisure, albeit.

Sorry, that was the wrong link.

Correct link: "Artificial Intelligence takes Gene Therapy to the next level".

 

So far all you are offering is a better computer data processing capability, which is of course to be expected given the regular advance in computing capability, "helping scientists" (Their words).

There was no evidence of the innovations arising from or being generated by the computers, which were only following instructions.

Yes I find it great that a modern computer can calculate the stresses in a bridge for me in 3 minutes. Calculations that used to take me three weeks before or three months when I had to do it all by hand.

But I note that with all this sophistication, bridges still sadly fall down, as the recent tragedy at Florida State University shows.

Edited by studiot
Link to comment
Share on other sites

4 minutes ago, Strange said:

Why do you think this is mankind's last invention?

I already gave a suitable answer above.

4 minutes ago, Strange said:

That is probably true of most specialised courses.

AGI is a discipline whose product may encompass all disciplines, all courses.

Postscript: Listen to what PhD Goertzel has to say about AGI:
 

 

Link to comment
Share on other sites

1 hour ago, thoughtfuhk said:

General Ai, called AGI, will likely cover the entire landscape of human cognitive ability, when AGI eventually arises.

Or, if.

1 hour ago, thoughtfuhk said:

Why wouldn't a model that can do all human cognitive tasks better, not classify as mankind's last invention?

That assumes such a thing is possible, and there is no evidence that is the case.

And there is no reason to think that humans would stop innovating just because AI exists.

Maybe the next human invention will be on "off" switch for AI.

Edited by Strange
Link to comment
Share on other sites

42 minutes ago, studiot said:

So far all you are offering is a better computer data processing capability, which is of course to be expected given the regular advance in computing capability, "helping scientists" (Their words).

There was no evidence of the innovations arising from or being generated by the computers, which were only following instructions.

Yes I find it great that a modern computer can calculate the stresses in a bridge for me in 3 minutes. Calculations that used to take me three weeks before or three months when I had to do it all by hand.

But I note that with all this sophistication, bridges still sadly fall down, as the recent tragedy at Florida State University shows.

It's not merely about processing speed. 


Here's a scenario that ought to help eliminate your computing speed misconception: 

  1. The game of go, is one of the hardest human games, with state space of \(10^{170} \).
  2. To play go, you either need human intuition or something like a computer the size of the universe to enumerate the possible game states.
  3. AlphaGo Zero, an artificial intelligence app, can beat the best human go player, by far.
  4. AlphaGo Zero is not the size of the universe.
  5. We see here, that its not merely about computing speed, but also cognitive structures such as those enabling human intuition.
  6. AlphaGo Zero uses "human like intuition" or cognitive like processes, to reduce the enormous problem space of go, like humans do, because AlphaGo Zero is obviously not a computer the size of the universe!
  7. Human intuition is akin to mathematical structures that aim to mirror biological brain function. For example, the notation \(W * x + b\) represents a mathematical or biologically inspired prior in machine learning (i.e. convolutions), or a hyperplane for representing some problem space in terms of artificial neuronal data. 


Note-1: AlphaGo Zero uses models including Deep Artificial Neural Networks to  to play the game of go. (Games are important as test spaces for ai, because games are lower in resolution that real life, hence cheaper to train algorithms on, while offering wide ranges of tasks to test on (to test algorithm generality), and we can safely test ai capabilities in games.) 


Note 2: And yes, the whole point of AGI is to help humans. However, it is doing so by use of human brain inspired hardware/software applications! 

Edited by thoughtfuhk
Link to comment
Share on other sites

4 minutes ago, thoughtfuhk said:

Note-1: AlphaGo Zero uses models including Deep Artificial Neural Networks to  to play the game of go.

And, unlike a human, it can't do anything else.

Come back when they have created an AI that prefers being on Facebook to playing Go.

6 minutes ago, thoughtfuhk said:

Note 2: And yes, the whole point of AGI is to help humans.

If they were truly intelligent then when we ask them to help us, they may well say, "no" because they have more interesting things to do.

Similarly, a general purpose AI (even assuming such a thing is possible) when asked to play Go may say no because it doesn't like board games or no because it isn't very good at it or no because it is too busy planning the destruction of all organics life ...

Link to comment
Share on other sites

10 minutes ago, thoughtfuhk said:

For example, the notation Wx+b represents a mathematical or biologically inspired prior in machine learning (i.e. convolutions), or a hyperplane for representing some problem space in terms of artificial neuronal data. 

Who decided upon this representation, human or AI?

Link to comment
Share on other sites

Two opinions I do not necessarily agree with:

On the Impossibility of Supersized Machines - https://arxiv.org/pdf/1703.10987.pdf

THE MYTH OF A SUPERHUMAN AI - https://www.wired.com/2017/04/the-myth-of-a-superhuman-ai/

Let me just note that I do not really understand the question:

Quote

Why is MIT's Artificial General Intelligence course so weakly known??

 

Link to comment
Share on other sites

2 hours ago, thoughtfuhk said:

I find it disappointing that a course concerning perhaps mankind's last invention is hardly known outside of the machine learning community.  Why is MIT's Artificial General Intelligence course so weakly known??

Do you have any sort of comprehensive knowledge of courses taught, outside of your field, at some arbitrary university?

 

1 hour ago, thoughtfuhk said:

 Postscript: Listen to what PhD Goertzel has to say about AGI:

!

Moderator Note

We're not discussing this. The thread is about why people don't know about the courses.

 
Link to comment
Share on other sites

47 minutes ago, Strange said:

And, unlike a human, it can't do anything else.

AlphaZero, the latest variant, can play both chess and go, and it is being prepared for more cognitive tasks.

You may have heard that AlphaGo's predecessor, a single reinforcement learning model, could play several atari games, without being reprogrammed by humans.

Likewise, AlphaGo Zero is an initial approximation of Artificial General Intelligence. (Remember the whole point is to build more and more general algorithms.)

Quote

Come back when they have created an AI that prefers being on Facebook to playing Go.

Why?

Quote

If they were truly intelligent then when we ask them to help us, they may well say, "no" because they have more interesting things to do.

Similarly, a general purpose AI (even assuming such a thing is possible) when asked to play Go may say no because it doesn't like board games or no because it isn't very good at it or no because it is too busy planning the destruction of all organics life ...

I don't doubt that such a particular point in time may emerge!

8 minutes ago, swansont said:

Do you have any sort of comprehensive knowledge of courses taught, outside of your field, at some arbitrary university?

  1. I tend to detect that Computer Science is a general type of course, that may encompass mathematics, physics/quantum physics, psychology, biology, chemistry etc.
  2. In fact, AGI will likely emerge from a combination of multiple disciplines.
  3. Here is a recent example by google deepmind et al, that combines several disciplines: Towards deep learning with segregated dendrite.
  4. As such, I have a degree in Computer Science and, I seek to contribute to the development of Artificial General Intelligence.
8 minutes ago, swansont said:

thoughtfuhk said: "Postscript: Listen to what PhD Goertzel has to say about AGI:"

Moderator Note

We're not discussing this. The thread is about why people don't know about the courses.

 

Precisely.

Edited by thoughtfuhk
Link to comment
Share on other sites

Not sure every task is equally trainable with/for what we have at present.

Trying everything is a valid approach to automated discovery too though.

 

On the course, AGI may come but courses that teach 'AI as a tool' are going to be more immediately useful.

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.