GPT-4 and Seed AI

[John Clark]

spudboy100 via Everything List <everyth...@googlegroups.com> wrote:

> My Trumpian amygdala has flat out taken my neocortex hostage. As we say in Trumpville, Golly!  While you are undertaking that arduous task, you never got a chance to address this impact question I submitted at the bottom of one of my rants. 

That's probably because you throw so many garbage links into almost every post you write I've gotten in the habit of ignoring them 

> This is an article, and I'd like you opinion of the likely impact of GPT-4 on research and development? I will re-supply the link.
https://towardsdatascience.com/gpt-4-will-have-100-trillion-parameters-500x-the-size-of-gpt-3-582b98d82253

This link was uncharacteristically interesting and reinforces my opinion that human level AI can't be very far away, and once that level has been achieved the attainment of superhuman AI can only take a few days more, perhaps only a few hours. I just don't see how we can be very far away from finding the seed algorithm that would allow computers to learn how to learn anything, because we already have an upper limit on how big that algorithm must be, and it's not very big. In the entire human genome there are only 3 billion base pairs. There are 4 bases so each base can represent 2 bits, there are 8 bits per byte so that comes out to just 750 meg, and that's enough assembly instructions to make not just a brain and all its wiring but an entire human baby. So the instructions MUST contain wiring instructions such as "wire a neuron up this way and then repeat that procedure exactly the same way 917 billion times". And there is a huge amount of redundancy in the human genome, so if you used a file compression program like ZIP on that 750 meg you could easily put the entire thing on a CD, not a DVD not a Blu ray just a old fashioned steam powered vanilla CD, and you'd still have room for a few dozen lady Gaga songs. And the thing I'm talking about, the seed learning algorithm for intelligence, must be vastly smaller than that, and that's the thing that let Einstein go from knowing precisely nothing in 1879 to becoming the first person to know General Relativity in 1915.

That's why I told you that whatever you do for a living it's only a matter of time, and not much time, before a machine can do your job better than you can. I always knew something like this would happen but I didn't think it would happen in my lifetime so I saw no need for an immediate change in my strict libertarian beliefs, but things have been developing much faster than I expected, and when new information becomes available an intelligent man must change his opinion. I'm still a libertarian with regard to social issues, but on economic matters not so much. It's just not rational to expect that the gargantuan and accelerating gap between the rich and the poor can continue indefinitely without causing a complete collapse of civilization. People just are not going to quietly starve to death, they are going to make a fuss. And that's also why I think all the worry Trump zombies have about Mexicans taking their jobs is ridiculous when the real problem is elsewhere.

John K Clark    See what's on my new list at  Extropolis

bo

[spudb...@aol.com]

I go with people who have proven trustworthy, rather than simply loyal like your Dem Media are to their pols. I do the same with Fox. 

On computation, I see much more impact than simply job loss as an impact. I look at the rapidity of progress outside the realm of quantum computing as amazingly fast, and it seems slower with QC. Medically we may be in for some amazing tech. Our big pharma friends are investing in GPT-4. I don't see Marvin Minsky's Guy in a Box and the best goal. I think just a fake intelligence that poops out innovations is totally fine with me! If I want to speak to a human level intelligence I will chat with my wife. If I want a better solar energy + battery system I will ask GPT-4. I am sure that it will be strictly for paying customers.  To this end here is a GPT 2+3/4 ths' mockup. 

Feel free to ask this network anything you like. I tend to go for the technological.  This is a semantic engine so play with the words and see if you get a interesting result from this computerized magic 8-Ball? Happy hunting...

https://6b.eleuther.ai/

All the best,

The Evil Trumpkin

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/everything-list/CAJPayv162dYfzeHxe0UUL1Wg%3DtWb68sNNGGnZYbu0caRZ54Amw%40mail.gmail.com.

[John Clark]

On Mon, Sep 20, 2021 at 1:50 PM <spudb...@aol.com> wrote:

https://6b.eleuther.ai/

It still needs work. I Think the ability to detect nonsense is important so I asked it "What is the oblong backstory of the refractory backhoe?" and instead of telling me my question was pure gibberish, as anyhuman would. it said this:

"The first thing I noticed in this picture is that the machine is parked on a raised platform at a right angle to the road. It looks as if it’s being lifted off the platform and will continue to be moved by crane. The second thing is that the backhoe has what looks like a refractory wall on top of the bucket, and there’s a steel ladder to the right. I looked it up and found the following: There are, however, two things not explained in the above picture. First, the backhoe was at a quarry where there is no bridge, so it was being lifted off the platform to be maneuvered over the gap. It had to be lifted onto the platform by crane because there was no bridge to the platform. Second, what was on top of the backhoe bucket was not a refractory wall. It was a hopper, in the front of which there were pipes for the refractory mixture to be poured from the hopper into the backhoe bucket. It was pouring the refractory mixture into the backhoe bucket, which was connected to a furnace, which was pumping it down into the face of the mine where the cutting was going on. And finally, what the ladder is for is to get to the top of the refractory mixture wall on the backhoe bucket to turn off the pumps and cut the refractory mix. How many horsepower were they putting out? As can be seen from the picture, there was a huge steel tube at the back of the refractory mixture wall. This tube was about 40 feet in diameter and had to be rotated by the backhoe. In the process of doing this, it was necessary to rotate the tube at least 20 revolutions per minute. This is called the “speed of the wheel”, and it is about 15}

It did even worse on my next nonsense question, I asked it "How many colorless green ideas slept furiously?" and it said this: "By the way of a first person account from the future, this was a very fun and successful experience, and it makes me optimistic about the future. I’ll start at the beginning. In the summer of 2013 I took the job as a grad student working with Tim O’Brien on a paper describing a new approach for solving a class of difficult optimization problems. The paper is about a certain kind of optimization called the “cut and project” problem: (I know, this seems horribly complicated. I’ll try to explain the basics in the next post.)"

John K Clark    See what's on my new list at  Extropolis

omx

[spudb...@aol.com]

Needs work? Yes, what we interacted with is a precursor to GPT-3 essentially, or GPT-3 de-rated. Rather than try to trip it up as a non-passing Turing test, I sort of did a couple of tries. One was a sci question, the other was the name of an ancient comedy album. I skimmed my own attempts from a small futurist forum and present these here as an example. It seems as if the obscure statement response was more meaningful in story-telling than the wormhole question was at analysis.

"I did just do this sentence however.
Don't crush that dwarf, hand me the plyers.

Response-

RESULT
Don't crush that dwarf, hand me the plyers. It's your first time working with dwarves, right? Good, don't let the blood splatter on your face.'

And then, in the time it takes to blink, the world turns upside down. The little guy's blood hits her face and he starts to flail around like a marionette. Then she hears the others' screams. She turns, and sees three of her fellow adventurers are running for their lives. They're being hunted by a pair of dwarven warriors and their gigantic dogs.

The last one runs at the dwarves, but the dogs have him in their jaws. He's going to be



and this one....
Can spaceships someday use cosmological wormholes for very fast travel?


Response-
RESULT
Can spaceships someday use cosmological wormholes for very fast travel? This week, at the 2017 International Astronautical Congress (IAC), a team of researchers presented some fascinating new results that demonstrate wormholes and wormhole throats can be used as spacetime tunnels for superfast travel.

“The main advantage of the new method we’ve introduced here is that it takes advantage of the already-existing physics of wormholes,” said Alessandro Nagar of NASA’s Goddard Space Flight Center. “That’s an important point: wormholes are there, so we can use them to our advantage.”

“There are a lot of things

-----Original Message-----
From: John Clark <johnk...@gmail.com>

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To view this discussion on the web visit

https://groups.google.com/d/msgid/everything-list/CAJPayv3zcM%3Dde99mVX8kKBeOF0sTepttG30odsNxzSriUJchcw%40mail.gmail.com

.