Groq CEO Jonathan Ross - Tech Giants in the Generative AI Age

[@Drackomass]

That was a really good episode, thank you for having it and sharing it. Jonathan Ross is really smart.

[@worldofdaas]

Glad you enjoyed it! Jonathan is amazing!

[@FutureGuy47]

Dude, stop interrupting him all the time.

[@kristopherleslie8343]

Agreed 😊

[@Superahh]

Yeah lol

[@Matrix1Gamer]

I appreciate Groq's business model for its chips because it is logical and makes perfect sense. Jonathan Ross makes great sense.

[@peter_phamous]

Really awesome guest! And while you're getting some warranted criticism on cutting off your guest too frequently, you asked some really excellent questions! But perhaps like Jonathan's point where LLM's, given more compute will produce much more beneficial responses, your questions can similarly benefit 😅. But well done, absolutely subscribed!

[@420_gunna]

Host is actually _good_ at interrupting in the sense that he seems to know the best times to kill interesting lines of thought from Groq man

[@SJtoobsox]

A hallmark of effective interrupting :)

[@andrewlewin6525]

Love the energy, but let him finish a thought 😂 but great stuff !!

[@worldofdaas]

Thanks for watching!! Please subscribe and share :)

[@frediv8]

Love how he is challenging the concept of data gravity.

[@videowatching9576]

Fascinating episode!

[@WorldMover]

Amazing

[@worldofdaas]

Thanks for watching. Subscribe to never miss an episode!

[@bstewartny]

You can't have compute without electricity, energy (oil, nat gas, coal, nuclear, etc). It's unlikely solar, wind and hydro can provide enough power in the short term for all these data centers full of GPUs. There will be huge incentive to make these GPU more energy efficient. And smaller more efficient models. And smaller models which can run on more efficient CPUs and not even need GPU necessarily (and even run locally on your own CPU - see LLama3 8B model for example)

[@worldofdaas]

Agree!

[@TheFeedRocket]

100% this, the models will get more efficient, smaller and smarter, by magnitudes, it will be possible to run and train AI multiple times better than ChatGPT on computers at home. We will look back at these massive fan and water cooled GPU's we use at home now as a joke, we still will have those, they will just be that much more powerful and cutting edge. We will run realtime extremely powerful AI models on computers at home in a year or so. They know that, and that is not their main goal really, the goal is to beat the other guy to super intelligence at all costs NOW. It's not about efficiency, not even about cost, it's just about winning the race, we have never seen anything like this. Lots of $$$ going into tech right now, things are accelerating more than ever on all fronts.

[@militiamc]

Interviewer keeps interrupting speaker just to ask basic questions or provide already obvious or sometimes flat out wrong examples. Regardless still a great interview.

[@worldofdaas]

is this negging?

[@thriceborn7665]

@@worldofdaas Not negging. It's real feedback. Good interview but would be great if you let this man cook lol. Best interviewers let their guest complete their thoughts without interruption.

[@worldofdaas]

@@thriceborn7665 that's fair. always appreciate feedback

[@thriceborn7665]

@@worldofdaas You got it man. I appreciate the work regardless. Great content here.

[@dimknaf]

stop interrupting. He was talking, and when he was about finishing to say the most important thing, you interrupt him. It makes the whole interview so annoying... He tries to come back to what he said to conclude and interrupt him again. Apart from this, great interview..

[@Iamguilherme]

Awesome ideas.

[@emetzger]

what's harder...building a competitive training chip or a competitive inferencing chip? if the former, what's to keep Nvidia from farting out a competitive inferencing chip and push out groq altogether?

[@mva6044]

2 tech bros, each of whom likes to hear himself talk.

[@worldofdaas]

tech brethren to you

[@mva6044]

@@worldofdaas touché. Though I wish the Q/A interaction included a bit more nuance, particularly about the downsides and current competitive landscape where the interviewer would press the guest not just take his talking points at face value. I know it's a tough line to tread, especially as one is growing a channel and wants more high profile guests -- Ezra Klein does a great job at this, but he's also got a staff to help and has hours & hours of reading & prep work for each interview. At the very least, I would strongly recommend reading "Nexus" by Yuval Harari, for an open-minded but still critical view of this developing technology.

[@worldofdaas]

@@mva6044 Thank you for such a thoughtful comment. Loved Sapiens. Will add Nexus to my to-read list

[@chirag2819]

Everyboding saying Auren came across odd, I guess he was just trying to clarify things.

[@worldofdaas]

Thanks for the nice comment - appreciate it :)

[@RussianQueenIrina]

He should have named it "reality potion" of Generative AI

[@raxcoins]

evga was the company nvidia bankrupted

[@jonnagap87]

What a terrible host! The host is clearly insecure and tries to prove to the guest he understands what he is being told by giving out unrelated and nonsensical analogies. Your follow up questions clear show that you are playing out of your league. Please let the guest completely present their idea instead of constantly trying to prove your capabilities!!

[@davidvasco]

which research found this claim? "when they paired a human doctor with the AI the uh results were worse than just having the AI give a diagnosis"

[@paultparker]

I don’t remember, but I did see the actual chart. Look at Google‘s medical LM research: med Gemini, or something like that. I think that that was where it was, although it could’ve been a paper on GPT4 in medicine.

[@delldoesai]

I still disagree with him about jobs. The distinction between coal engines and pace of change as he initially mention of AI adoption is significantly fsster. Slow change is adaptable, fast change requires a long time to balance out frictional employment.

[@eatdirtnetwork]

The fact that he was dishonest w the last 2 questions proves he is a person disconnected w reality

[@cheese9812]

worst interviewer ever. just interrupts his guest the entire time. hes as bad as kara swisher

[@ps3301]

Groq is trying to sell itself to apple..groq is using Samsung foundry for the next gen chip. Not north American supply anymore

[@markwvh]

Actually, the Samsung foundry is based in Taylor, Texas, USA.

[@user-cv2as4jo9l]

Great choice of Samsung over TSMC. GL XD

[@mickelodiansurname9578]

Right now Groq's main use case, speed, is being throttled by Groq due to demand... But I suppose if you are just building and testing its okay... still really annoying though.

[@frankwj73]

Who cares!!!!!