15 Comments
User's avatar
Filip L's avatar

I´m working in the industry and predictions is we gonna hit high constraints in 2026 with chips as well, so it´s gonna be high demand on hardware on hand. Kinda like the covid shortage only on steroids.

Every one wants them and geopolitics are moving in to secure it as well

Jordamøn's avatar

We have a piece coming Monday on that.

J Scott's avatar

The protected thing will become "good workflows."

The best ones will be like the old ones, requiring capable humans..

Currently used it for:

Grading first pass.

Book notes.

Voice to prose.

Verse adaption of prose.

Image generation.

Newest based on Vox's influence:

Translation.

So far, 99% local models.

Shefi1280's avatar

“Most GenAI systems don’t retain feedback, adapt to context, or improve over time. They repeat the same mistakes and require extensive context input for each session.”

So they don’t learn. They are not intelligent, just artificial. And in other news, water is wet!

Chris's avatar

"most systems"

Are you prepared to admit that the minority of systems who do retain feedback, adapt to context, and improve over time are in fact intelligent?

Vox Day's avatar

AI is ABSOLUTELY worth it from my perspective. It has literally supercharged my personal productivity as an author.

And considering that all of them have gone to #1 in their categories, some quite large, such as Biology, it's impossible to dismiss them as AI slop vampire harem nonsense.

Snowyteller's avatar

Unfortunately, suits, being headless chestless beings of expensive fabric, do not possess such a baleful and unshaken pair of eyes to hold perspective true.

They don't even have eyeballs after all.

Still, powers great and small, fair and foul are quite committed to the integration being done, whatever the masses, the suits or whoever else think.

Well, we can look to history at how it goes with game changers, you'd certainly know far better than this teller.

GH's avatar

AI is the real IQ proof.

Results from the 115s just dont look like results of a 160.

It matches the pattern and generates.

Jordamøn's avatar

Indisputably; you're one of the 5%.

Claude Code's $500M in revenue last year mainly comes from similar use by individuals and small teams who figured out how to make it work for them.

It may be that's where the real value is accumulating.

Avalanche's avatar

Vox is "one of the 5%."

"It may be that's where the real value is accumulating."

The real value was already present before AI showed up. The AI could NOT have done any of this without Vox to start, steer, question, stop, restart, steer and question some more, redirect, and finally polish the amazing work he had done/is doing WITH this double-super-fake-smart calculator. (No offence intended to Claude Athos.)

I have an ongoing delight envisioning driver Vox Day in his super-high-performamce Fornula One racing car, screaming around the Monaco track, producing fantastic results both in speed and handling (and smiling the whole time!) He has tweaked and upgraded his 'car' and 'taught it to do stuff' it does not natively have.

Without Vox? It's not much more than car. The car with some OTHER driver? Nothing special. And apparently there are MANY other drivers, spending other people's $gazillions and not getting any kind of decent lap-times or paybacks. The "special" came in with Vox.

The tool is superb in the hands of a master. How many MASTERS are out there? The world-wide value of AI will absolutely require having masters or at the very least high-level apprentices ABLE to drive these racing cars and not drive them off into the woods!

Vox recently wrote about one his Ilk finding a number-glitch in a (was it 2019?) paper of Vox's... and Vox was ABLE TO 'drive' his co-author Claude to work it out with him. But Claude did not find it. Vox steered him to it. (The MASTER, not the machine!)

The average Indian IQ is in the 70s. 75 is the boundary for retarded. EVEN IF 'they're sending us their best" -- one might look askance: their best grifters? -- that is in NO way sufficient to drive a Formula One racing AI.

Jordamøn's avatar

Exactly. We're not talking about the value of Vox here, that had already been demonstrated beyond any shadow of a doubt long before GenAI became the phenomenon it is today.

People like Vox are the supremely talented racing driver who for one reason or another never got put behind the wheel of a championship-caliber car; maybe they got hired by bad teams, maybe they didn't get hired at all, but they proved themselves time and again in lesser categories.

Now, with AI, they're being given the car, the strategist, the mechanics, all for pennies on the dollar.

Avery Mann's avatar

It might be worth taking a gander at what's going on at https://fractalcomputing.substack.com/

Interesting developments on the "efficiency" side of things . . .

Teleros's avatar

I use AI for writing & research, and it's not close between Claude & ChatGPT - the latter is just flat-out worse. ChatGPT can still do images, and the Deep Research tool is useful, but those are the only reasons I maintain a subscription.