Meet Transformers: The Google Breakthrough that Rewrote AI's Roadmap

[Julio Franco]

For all the talk about AI these days, you may not know that it was a 2017 Google research paper that kickstarted modern generative AI by introducing the Transformer, a groundbreaking model that reshaped language processing.

Read the full article here.

Meet Transformers: The Google Breakthrough that Rewrote AI's Roadmap

 

[WhoCaresAnymore]

Google's AI roadmap is built on baked in woke programming and misinformation. Most people I know here in the US don't use Google anymore. They are also looking for gmail alternatives for the long term.

Google isn't the only problem with misinformation, the general media is just in shambles.

 

[FaTaL]

So the AI breakthrough was give computers adderall to focus. Got it.

 

[DekkB]

The giant robot has three legs and what looks like an old car exhaust sticking out of his head. Err?

 

[Ashford]

I'm waay to deep into the nog to wade into this tonight ... but this was supposed to be my escape from an awkward family conversation....

Dammit... they told me to out my phone away...

 

[Nearchus]

Having first used expert systems (inference engines, rule chaining) back in the late 80's and encountering neural nets then, it is interesting how much the math and algorithms have refocused on treating the data as a sea of numbers and letting coherence emerge, so to speak, rather than dictating its direction. Babies, of course, quietly get on doing it without fuss or bother and later emerge as some ***** playing COD.

 

[Porkous]

The key aspect is that the performance scales with more datasets.

I'm still not seeing it as the future, unless AI chips become more common, and are not just designed by the most profitable company, just because it can.
If you notice, Nvidia is not looking to cut prices and make affordable products, au contraire.
So this scalability, not having enough chips, I think, is what is truly stopping the advance.

Also ironically, these big companies having to spend so much, as if AI right now is synonym with big capital investments, and not abundance of availability, itself... Maybe this slow down aspect is good for us... As it holds the power of AI in leash, until people truly know what to do with it.

The datasets are free? So the very essence of AI should be availability... not being locked by big capital expenditure of these companies.

I like that the whole transformer architecture is based on a research, brainstorm from scientists.

 

[airbornex]

The giant robot has three legs and what looks like an old car exhaust sticking out of his head. Err?

LoL. I think that was like a graphical form of "pun intended" by the author.

 

[ScottSoapbox]

Google's AI roadmap is built on baked in woke programming and misinformation. Most people I know here in the US don't use Google anymore. They are also looking for gmail alternatives for the long term.

Google isn't the only problem with misinformation, the general media is just in shambles.

At least read the article and make your political rant somewhat related to it. This a tech site not Twitter.

 

[billc]

Having first used expert systems (inference engines, rule chaining) back in the late 80's and encountering neural nets then, it is interesting how much the math and algorithms have refocused on treating the data as a sea of numbers and letting coherence emerge, so to speak, rather than dictating its direction. Babies, of course, quietly get on doing it without fuss or bother and later emerge as some ***** playing COD.

The key aspect is that the performance scales with more datasets.

I'm still not seeing it as the future, unless AI chips become more common, and are not just designed by the most profitable company, just because it can.
If you notice, Nvidia is not looking to cut prices and make affordable products, au contraire.
So this scalability, not having enough chips, I think, is what is truly stopping the advance.

Also ironically, these big companies having to spend so much, as if AI right now is synonym with big capital investments, and not abundance of availability, itself... Maybe this slow down aspect is good for us... As it holds the power of AI in leash, until people truly know what to do with it.

The datasets are free? So the very essence of AI should be availability... not being locked by big capital expenditure of these companies.

I like that the whole transformer architecture is based on a research, brainstorm from scientists.

Scientists who don't understand that quantum mechanics plays a big role in a brain, and is not at all well understood? All the engineers who left Altman didn't do so because that thought AI might become intelligent, rather because they know he is talking out of his arse, and were tired of being mocked by friends, and so decided to look for more realistic jobs.

 

[ccccn]

Google's AI roadmap is built on baked in woke programming and misinformation. Most people I know here in the US don't use Google anymore. They are also looking for gmail alternatives for the long term.

Google isn't the only problem with misinformation, the general media is just in shambles.

Dead on accurate , it's a shame many still don't have a clue about this.

 

[JaredTheDragon]

And not one shred of this entire topic, process, or algorithm is remotely "intelligent". It's just so silly that anyone fell for ANY of this! And insane.

 

[JaredTheDragon]

Scientists who don't understand that quantum mechanics plays a big role in a brain, and is not at all well understood? All the engineers who left Altman didn't do so because that thought AI might become intelligent, rather because they know he is talking out of his arse, and were tired of being mocked by friends, and so decided to look for more realistic jobs.

What QM effects occur in the human brain, though? Almost all of QM, QCD, and QED has been falsified in the last two decades, including superposition, entanglement, tunneling, quarks, and everything Heisenberg, Bohr, and Feynman ever posited.

What is the quanta you're even referring to? You cannot say. What is electricity? You do not know. What about magnetism? Not a chance.

Leave physics to the physicists, or join us through rigorous study.

 

[billc]

What QM effects occur in the human brain, though? Almost all of QM, QCD, and QED has been falsified in the last two decades, including superposition, entanglement, tunneling, quarks, and everything Heisenberg, Bohr, and Feynman ever posited.

What is the quanta you're even referring to? You cannot say. What is electricity? You do not know. What about magnetism? Not a chance.

Leave physics to the physicists, or join us through rigorous study.

But what about the five sided quasicrystal?