MVissers

How? Interesting co’s are private (openai, groq, etc).

Nvidia already huge. Microsoft and Apple are more users.

Probably within 5 years. Compute growing exponentially, algorithm improvement as well, multi modalities allows for different type of data training, etc…

Yeah, this shit is near. Also- Quite a dangerous experiment we’re running. And safety-first people are not at the helm anymore.

I believe GPT-4 pre-rlhf was much more accurate overall.

If you finetune it in formats that humans likes, it actually gains similar biases.

I believe the ‘sparks of AGI’ talks about this, where models can much more accurately predict the probability of events than humans.

After RLHF, it mimicks human bias. So it might be that we can create these models, but we just don’t like them/like to use them.

Good comparison.

There is more money in AGI than LLMs.

Whatever it is, language seems key to intelligent algorithms.

Nah, he departed due to politics (failed coup) and shift from research first to profit first. Same with Karpathy I believe.

He’ll most likely go somewhere where he can get a lot of compute and go back to research first.

Nah– These models will improve faster than people can catch up. People or AI models can barely catch AI-created text. It's quickly becoming impossible to distinguish.

The one you catch is the tip of the iceberg.

Same will happen to speech. Might take a few years, but it'll be indistinguishable in a max a few years. Due to compute increase + model improvement, both improving exponentially.

Soon these models are cheap enough to learn in the real world. Reduced costs allows for usage at massive scale.

Releasing models to users that where users can record video is more data. Users conversing with AI is also additional data.

Another example is models that code– And then debug the code and learn from that.

This will be anywhere, and these models will learn from anything we do/publish online/discuss. Scary.

Pretty soon– OpenAI will have access to

Why would reasoning have plateau’d?

I think reasoning ability is not the largest bottleneck for improvement in usefulness right now. Cost is a bigger one IMO.

Running these models as agents is hella expensive, and agents or agent-like recurrent reasoning (like humans do) is the key to improved performance if you look at any type of human intelligence.

Single-shot performance only gets you so far.

For example- If it can write code 90% of the way, and then debug in a loop, it’d be much more performant than any single shot algorithm.

And OpenAI has these huge models in their basement probably. But they might not be much more useful than GPT-4 when used as single-shot. I mean, what could it do what we can’t do today with gpt-4?

It’s agents and recurrent reasoning we need for more usefulness.

At least- That’s my humble opinion as an amateur neuroscientist that plays around with these models.

They did the same with GPT-4, they were sitting on it for months not knowing how to release. Ended up releasing GPT-3.5 and releasing 4 quietly after nerfing 3.5 into a turbo.

OpenAI sucks at naming though. GPT2 now? Their specific gpt-4-314 etc. model naming was also a mess.

In Europe they have to refund your ticket plus compensate you in cash depending on delay and distance.

Not sure if this will change much in the USA, refunding is not that high of a cost either for airlines.

There is now way this is overblown hype. This is real- And the progress on AI capabilities is exponential with no signs of slowing down.

Agree that we don’t think enough about it.

Stop paying for chatgpt and use platform.openai.com instead to pick your model for consistency.

I’m still using the original gpt-4-3.14 model.

You could- That’s the case in Europe. But ‘Murica = Freedom so no interest to regulate.

Forests that never decayed because nothing could break down liglin molecules for millions of years. Many buried underground and I believe turned into coal/oil.

Do you have data for that?

Occupancy is at record highs in most US cities. It’s simply too little inventory, mostly due to local regulations and NIMBY’s. Most housing is owned by ppl, and a lot of the nation’s net worth is tied up in housing, disencentivising the lowering of house pricing.

You could see it duting covid in NY, people left and inventory went up. Rent and housing came down.

Same now in SF, ppl leave, inventory goes up, prices come down.

It’s not that complex really. Solving it will not be easy though.

What’s also fascinating is that a child’s brain has additional plasticity/aptitude for learning different things at a different age.

If he/she learns it too late, they’ll never master it. Perfect pitch is an example, ‘perfect’ motor skills for something another. Language another.

Finetuning happens as a teenager/adult.

“The market can stay irrational longer than you can stay solvent.”

Eventually- Markets go back to the mean, but predicting when is the hard part.

We have regions that specialize in specific tasks. Information gets routed there for recognition and understanding.

Eg: Faces are processed in the fusiform area. And if you play pokemon obsessively as a kid, you’ll even create an expert pokemon region: https://news.stanford.edu/2019/05/06/regular-pokemon-players...

Depends. Unchecked it’s completely disastrous to most except a few.

I do believe we should get more oversight and have co’s invest more in research on safety/alignment.

It’s a wild west right now with potentially fatal end result to humanity.

I mean, basically no-one can compete with OpenAI right now.

They have a monopoly. They can start copying any business that relies on them tomorrow.

And it’s not easy to build a better model. Do that’s not an option.

So you basically use them to deliver a tangential product woth the risk of them copying you, or refuse to play in AI.

You can’t innovate with OpenAI tech without risking that they’ll just copy you

It’s the amazon playbook: become the store and copy the best selling products.

I love gpt-4, just hate the direction of openAI. Sam knows he can take over the world with OpenAI, so he’s just doing it.

This is the scientific consensus btw.

There are also more intellectually challenged men btw, but somehow that rarely gets discussed.

But the effects are quite small, and should not dissuade anyone to do anything IMO.

“And that moment was the final nail in the coffin of humankind from earth. They choose, yet again, for money and power. And they shaped AI in their image.

Another civilization perished in the great filter.”

They’ll all be filthy rich if they can keep doing this. Altman was already side-hussling to get funding for other AI companies.

Same with employees and their stock comp. Same with microsoft.

If by “long-term-success” you mean a capitalistic lap-dog of microsoft, I’ll agree.

It seems that the safety team within OpenAI lost. My biggest fear with this whole AI thing is hostile takeover, and openAI was best positioned to at least do an effort to prevent that. Now, I’m not so sure anymore.

Let’s not creating AI with our biases and thought patterns.

Oh wait…

As long as compute keeps increasing, model size and performance can keep increasing.

So no, we’re nowhere near max capability.

GPT-4 on platform.openai.com says this on the first try:

Switch on the first switch and leave it on for a few minutes. Then, switch it off and switch on the second switch. Leave the third switch off. Now, walk into the room.

The bulb that is on corresponds to the second switch. The bulb that is off and still warm corresponds to the first switch because it had time to heat up. The bulb that is off and cool corresponds to the third switch, the one you never turned on.

GPT-4-0314: 1. Turn on the first switch and leave it on for about 5 minutes. 2. After 5 minutes, turn off the first switch and turn on the second switch. 3. Open the door and enter the room.

Now observe the lights: - The bulb that is on is connected to the second switch (which is currently on). - The bulb that is off but warm to the touch is connected to the first switch (it was on long enough to heat up the bulb). - The bulb that is off and cool to the touch is connected to the third switch (it was never turned on).

----

But– It's also trained on the internet. GPT-4 paper 'sparks of AGI' had a logical puzzle it most likely never encountered in the training data that it could solve.

Also– I encourage you to ask these types of logical puzzles on the street to rando's. They're not easy to solve.

My question to you would be: What would convince you that it actually can 'think' logically?

So when your drunk and you forget your actions the day afterwards you don't consider yourself to have been sentient/conscious?

That's not how we define conscious anywhere.

You can process the world around you, feel and introspect. Even if your judgment is off and you forget your actions, you're conscious in that moment.

From a neuroscience perspective, what you're suggesting is absolutely false btw.

The board of the non-profit (one that fired Sam) has no fiduciary duty to those investors, I believe. Microsoft invested in the for-profit Openai, which is owned by the non-profit. The other ones I don't know.

The board has no responsibility to Microsoft whatsoever regarding this. Sam Altman structured it this way himself. Not to say that the board didn't screw up.

You should read the safety paper of GPT-4. It can easily manipulate humans to attains it goals.