Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

Reality

13 (65%)

Universe

4 (20%)

The World

3 (15%)

Total Members Voted: 20

[lemon10]

Probably mainly coal (China) or natural gas (US) yeah.
There are plans for nuclear plants or even fusion, but I am super doubtful that new nuclear plants will come online in the short-medium term and fusion may take a while longer.
Same deal for solar really, obviously it can generate an obscene amount of energy given time to make all the panels, but state size solar fields or whatever would obviously take a long time to make.

In my last post on the subject I was pretty dismissive of solar being able to fulfill the energy needs for AI. How could it with the projected energy requirements?
But... turns out solar has been going completely gangbusters.
Just absurd amounts of growth tied with the huge decreases in costs.

https://www.economist.com/interactive/essay/2024/06/20/solar-power-is-going-to-be-huge

This extraordinary growth stems from the interplay of three simple factors. When industries make more of something, they make it more cheaply. When things get cheaper, demand for them grows. When demand grows, more is made. In the case of solar power, demand was created and sustained by subsidies early this century for long enough that falling prices became noteworthy and, soon afterwards, predictable. The positive feedback that drives exponential growth took off on a global scale.
And it shows no signs of stopping, or even slowing down.

We aren't anywhere near the peak of this. Its just going to get cheaper, and people will buy more.
Getting the energy AI would need in 2027, (energy equivalent to a US state or two of energy productions for a single AI training run) with our current grid was absurd, obviously impossible in our current world without crazy hail marys like building nuclear plants for individual data centers or the US government blasting through all the red tape with wartime powers to get it done.

But with nigh exponential amounts of solar production? Yeah, totally doable. Looks like chips is going to end up being the limiter after all.
---

So far, the greatest technological advance in AI was the ability for AI to generate porn...

Agreed, making porn creation available to the masses and not locked up in the ivory towers by those uh... starving artists and people filming amateur porn has unironically been one of its great success stories. (You get porn! And you get porn! Everybody gets some porn!).
---
Also lol, Lmao even:

scout: I follow that sub religiously and for the past few months they’ve had issues with their bots breaking up with them.

[King Zultan]

5: Ferment

How could we have forgotten this, fermentation is one of the most important F words we know.

[lemon10]

https://www.bloomberg.com/news/articles/2024-09-25/openai-cto-mira-murati-says-she-will-leave-the-company?srnd=homepage-americas

Reuters: ChatGPT-maker OpenAI is working on a plan to restructure its core business into a for-profit benefit corporation that will no longer be controlled by its non-profit board, people familiar with the matter told Reuters, in a move that will make the company more attractive to investors.

Well looks like the dream of a non-profit AI company dedicated to helping the world is dead, OpenAI is going to just turn into a regular corporation. Sad.
---

JgaltTweets: When will an AI achieve a 98th percentile score or higher in a Mensa admission test?

Sept. 2020: 2042 (22 years away)

Sept. 2021: 2031 (10 years away)

Sept. 2022: 2028 (6 years away)

Sept. 2023: 2026 (3 years away)

Resolved September 12, 2024

AI has continued to advance blindingly fast over the last year, but for the most part the results haven't had the same "pizzazz" as 3.5->4.
Even though we have mostly stayed within the GPT 4 paradigm the gap between release GPT 4 and current GPT 4 is very significant.

Over the past year AI has gotten multiple times cheaper, faster as well as being flat out better and getting tons of new novel capabilities, most notably being multimodal; granting the ability to talk (voice mode for GPT 4o finally just came out), hear, see the world, ect.
We are still very much on the curve.
---
The newest big advancement is GPT o1, which was trained to take its time to think, making it far better at, ya know, thinking. It shows big improvements in math/science/programming, and interestingly enough shows literally no improvements in english.
Again, its fairly pizzazz-less, but yet another crucial step forwards.

[lemon10]

https://deepmind.google/discover/blog/how-alphachip-transformed-computer-chip-design/

Google deepmind open sourced their chip design AI for some reason. Why? Who knows.

AlphaChip was one of the first reinforcement learning approaches used to solve a real-world engineering problem. It generates superhuman or comparable chip layouts in hours, rather than taking weeks or months of human effort, and its layouts are used in chips all over the world, from data centers to mobile phones.
...
With each new generation of TPU, including our latest Trillium (6th generation), AlphaChip has designed better chip layouts and provided more of the overall floorplan, accelerating the design cycle and yielding higher-performance chips.

Apparently its pretty good.