Re: [agi] GPT-4o

[immortal . discoveries]

On Tuesday, May 14, 2024, at 10:27 AM, Matt Mahoney wrote:
> Does everyone agree this is AGI?
> 
It's not AGI yet because of a few things. Some are more important than others. 
Here is basically all that is left:

It cannot yet do long haul tasks that take weeks, and much steps. Ex. create 
Windows 12.

It cannot yet learn online very fast, only in monthly batches or with a limit 
aim for network size. I guess that's how to say it? Correct me if understand it 
wrong.

It has no body integrated.

No video AI integrated.

And they said in the email it is as smart as GPT-4 Turbo I tried, which failed 
my hard puzzle as bad as early GPT-4. My secret hard puzzle is not overly 
large, it says to stick to physics and gives it a dozen things to 
combinationally use and pick between. It is a mind-bending test to hell that is 
simple enough as hell that a human should know how to solve it in the room and 
setting provided. GPT-4 instead says things like it will use the spoon to 
tickle out the water from the other side of the room to get the gate to come 
down, and that it can sneak by the cloud and ask it to leave even though I said 
it cannot talk and does it's thing stated, for the cloud.
--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T17fa3f27f63a882a-M32c35511ce81608b8118367e
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Iteratively Tweak and Test (e.g. MLP => KAN)

[Matt Mahoney]

The top entry on the large text benchmark, nncp, uses a transformer. It is
closed source but there is a paper describing the algorithm. It doesn't
qualify for the Hutter prize because it takes 3 days to compress 1 GB on a
GPU with 10K cores.

The winning entry, fx-cmix, is open source. It is a variation of cmix,
which uses the PAQ architecture that I developed. It has a lot of
independent bit predictors whose predictions are combined using a simple 2
layer neural network. A prediction p is stretched as x = ln(p)/ ln(1-p).
The output prediction is squash(sum_i xi wi) where w is the weight vector
and squash(x) = 1/(1+e^-x) is the inverse of stretch. The weights are then
updated by w = w + L(y-p) where y is the actual bit, p was the prediction,
and L ≈ .001 is the learning rate.

You can find the software, algorithm descriptions and benchmark results at
https://mattmahoney.net/dc/text.html

For more about data compression in general, including the PAQ algorithms,
see
https://mattmahoney.net/dc/dce.html


On Sun, May 12, 2024, 9:14 PM John Rose  wrote:

> On Sunday, May 12, 2024, at 10:38 AM, Matt Mahoney wrote:
>
> All neural networks are trained by some variation of adjusting anything
> that is adjustable in the direction that reduces error. The problem with
> KAN alone is you have a lot fewer parameters to adjust, so you need a lot
> more neurons to represent the same function space. That's even with 2
> parameters per neuron, threshold level and steepness. The human brain has
> another 7000 parameters per neuron in the synaptic weights.
>
>
> I bet in some of these so-called “compressor” apps that Matt always looks
> at there is some serious NN structure tweaking going on there. They’re open
> source, right? Do people obfuscate the code when submitting?
>
>
> Well it’s kinda obvious but transformations like this:
>
> (Universal Approximation Theorem) => (Kolmogorov-Arnold Representation
> Theorem)
>
> There’s going to be more of them.
>
> Automating or not I’m sure researchers are on it.
> *Artificial General Intelligence List *
> / AGI / see discussions  +
> participants  +
> delivery options 
> Permalink
> 
>

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T1af6c40307437a26-Ma01352c6397139afc00fd032
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] GPT-4o

[Matt Mahoney]

On Tue, May 14, 2024, 11:23 AM James Bowery  wrote:

> Yet another demonstration of how Alan Turing poisoned the future with his
> damnable "test" that places mimicry of humans over truth.
>

Truth is whatever the majority believes. The Earth is round. Vaccines are
safe and effective. You have an immortal soul. How do you know?

I agree that compression is a better Intelligence test than the Turing
Test. But Intelligence is not the goal. Labor automation is the $1
quadrillion goal. The Turing Test is a check that your training set is
relevant.


--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T17fa3f27f63a882a-M2a4e691c2001ae18dd082537
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] GPT-4o

[Matt Mahoney]

AI should absolutely never have human rights. It should be illegal for an
AI to claim to be conscious or have feelings. ChatGPT already complies. I'm
pretty sure most other AIs do too.

We build AI to serve us, not compete with us. Once it does that, it wins.
The alignment problem is how to prevent this.

An AI predicts human actions. If you program it to carry out those
predictions in real time, then it passes the Turing Test and appears to be
conscious and have feelings as far as you can tell. But an AI can be
programmed to do other things with those predictions that you can't, and
you are already seeing the results.


On Tue, May 14, 2024, 12:55 PM  wrote:

> The question that really interests me is: what would GPT-4o say and do if
> given rights equal to human rights. I feel there is a lot more potential in
> the technology other than to blindly follow our instructions. What I want
> to see is some critical opinions and actions from AI.
> *Artificial General Intelligence List *
> / AGI / see discussions  +
> participants  +
> delivery options 
> Permalink
> 
>

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T17fa3f27f63a882a-M83f4b75b17897e941fa93354
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Ruting Test of AGI

[Keyvan M. Sadeghi]

where the slugs are called "riders".

Better ride than be ridden, especially when fuckers like Altman are driving
the world!

In his below interview, he outsources the worries, despite being the only
person in the world currently in possession of resources to address the
said worries:

https://www.businessinsider.com/sam-altman-says-ais-economic-impact-top-concern-2024-5

It doesn’t hurt to at least have a clue of how your product works when one
is the CEO of a behemoth!

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T244a8630dc835f49-Mb24144659abbf8e689f64009
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Ruting Test of AGI

[James Bowery]

Currently reading "The Puppet Masters" where the slugs are called "riders".

On Tue, May 14, 2024 at 11:21 AM Keyvan M. Sadeghi <
keyvan.m.sade...@gmail.com> wrote:

>
>
> That you find "tyranny for the good of their victims" "philosophical"
>> rather than "direct" indicates your ethical poverty.
>>
>
> More wise words from under the blanket ;)
> *Artificial General Intelligence List *
> / AGI / see discussions  +
> participants  +
> delivery options 
> Permalink
> 
>

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T244a8630dc835f49-M8bdbd808403c6e9eb4cbc2dd
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] GPT-4o

[ivan . moony]

The question that really interests me is: what would GPT-4o say and do if given 
rights equal to human rights. I feel there is a lot more potential in the 
technology other than to blindly follow our instructions. What I want to see is 
some critical opinions and actions from AI.
--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T17fa3f27f63a882a-M97d49733e8ce458a89d5d1dd
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Ruting Test of AGI

[Keyvan M. Sadeghi]

That you find "tyranny for the good of their victims" "philosophical"
> rather than "direct" indicates your ethical poverty.
>

More wise words from under the blanket ;)

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T244a8630dc835f49-M26f5902a39bef67b4e2fa191
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Ruting Test of AGI

[James Bowery]

That you find "tyranny for the good of their victims" "philosophical"
rather than "direct" indicates your ethical poverty.

On Tue, May 14, 2024 at 8:20 AM Keyvan M. Sadeghi <
keyvan.m.sade...@gmail.com> wrote:

> The Sam Altmans of the world are bound and determined to exercise tyranny
>> for the good of their victims -- which amplifies any mistakes in choosing a
>> world model selection criterion (ie: loss function).
>>
>
> Too philosophical for my taste, I like being direct and express my
> feelings in real world:
>
> https://x.com/keyvanmsadeghi/status/1790369335153742081
> 
>
> > @bengoertzel is a hero who started his crusade against bigotry when my
> generation were infants. The world owes him the scientific foundation of
> #AGI, that for the time being is represented by capitalist zealots like
> @sama.
>
> *Artificial General Intelligence List *
> / AGI / see discussions  +
> participants  +
> delivery options 
> Permalink
> 
>

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T244a8630dc835f49-M7f4b7e5e743222663563be0a
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] GPT-4o

[James Bowery]

Yet another demonstration of how Alan Turing poisoned the future with his
damnable "test" that places mimicry of humans over truth.

On Mon, May 13, 2024 at 3:55 PM  wrote:

> https://openai.com/index/hello-gpt-4o/
>
> Human voice finally, can be told to talk faster and can laugh and sing etc.
>
> It also has advanced image generation, see the examples.
>
> It seems to be maybe GPT-4.5 or GPT-5 also. Still checking it out.
>
> Coming to chatGPT in upcoming weeks.
> *Artificial General Intelligence List *
> / AGI / see discussions  +
> participants  +
> delivery options 
> Permalink
> 
>

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T17fa3f27f63a882a-M281d496f7a63d0cf2db03572
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] GPT-4o

[Matt Mahoney]

Does everyone agree this is AGI?

>From the demos it seems to be able to do all the things a disembodied
human can do. Although I saw on Turing Post that the public version can't
sing or stream video.

On Mon, May 13, 2024, 4:55 PM  wrote:

> https://openai.com/index/hello-gpt-4o/
>
> Human voice finally, can be told to talk faster and can laugh and sing etc.
>
> It also has advanced image generation, see the examples.
>
> It seems to be maybe GPT-4.5 or GPT-5 also. Still checking it out.
>
> Coming to chatGPT in upcoming weeks.
> *Artificial General Intelligence List *
> / AGI / see discussions  +
> participants  +
> delivery options 
> Permalink
> 
>

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T17fa3f27f63a882a-Mb2f8f903ca8d2467fd0bdb9c
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Ruting Test of AGI

[Keyvan M. Sadeghi]

>
> The Sam Altmans of the world are bound and determined to exercise tyranny
> for the good of their victims -- which amplifies any mistakes in choosing a
> world model selection criterion (ie: loss function).
>

Too philosophical for my taste, I like being direct and express my feelings
in real world:

https://x.com/keyvanmsadeghi/status/1790369335153742081


> @bengoertzel is a hero who started his crusade against bigotry when my
generation were infants. The world owes him the scientific foundation of
#AGI, that for the time being is represented by capitalist zealots like
@sama.

--
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T244a8630dc835f49-M381e37dd5d82fe81c6cd0b29
Delivery options: https://agi.topicbox.com/groups/agi/subscription