Adrian Kaehler

Join me on a tour and a look at what we at @GiantAITalks have been working on.

I hear @reidhoffman just turned a short blog post into a 230 page book using AI. Everyone has been mass producing gibberish on the internet for so long, all they can think to do with AI is produce more gibberish. The value in AI is in reducing the noise, not in amplifying it.

@claughlin @kittukolluri @nealkhosla @vkhosla Kudos for keeping it real. Still, credit where credit is due. Good investors structure their investment and their portfolios so that their interests are aligned with their companies, and stay aligned, even when things get a little strange.

Some VCs are notably absent and not bridging their portfolio companies right now. @vkhosla took out a PERSONAL line of credit to bridge at borrowing cost to his. Think really hard about who you take money from. And if you’re not stepping up as a VC right now, be better.

Am I the only one who has noticed that #chatcpt responds better when you say please?

@DrHughHarvey You've read Thinking Fast and Slow? There is also Buying Fast and Slow. The fast system makes fast buying decisions, it's easily influenced by bullshit. The slow thinking system avoids BS, but decides slower, and buys slower. Of course, it also writes bigger checks...

@araskin You had expected otherwise?

Series C CEO told me recently remote work is killing productivity mostly because people aren't putting in the hours

@GaryMarcus There is nothing unreasonable about this. Using web aware LLM's like GopherCite/Sparrow, LaMDA, or WebGPT (etc.) Is an incredibly natural way to retrieve information. Most of us didn't even notice that we had to build a pretty sophisticated skill to use search engines effectively

Enthusiast: ChatGPT is gonna replace search engines like Google! Skeptic: Yeah, but it doesn’t really work. sometimes it is amazing, but it often gives you garbage. Enthusiast: Sure but you can make it work! All you have to do is … hook it up to … a …search engine! 🙄

#ChatGPT is amazingly effective at inducing Twitter posts.

@_jasonwei @LukeZettlemoyer Keep up the good work!

@zackkanter It's actually quite a challenge to find people with the skill set a board requires and real technical knowledge. I think this reality very much is underappreciated. This becomes even more true for highly technical "deep tech" companies.

This is a really interesting point. Has anyone actually been inside a Tesla “factory,” or even seen one drive on a public road (not parked on the street, but actually driving, with your windows down / close enough to hear any giveaway sounds)?

@Scobleizer @samanthagwolfe @_LucasRizzotto What am I supposed to be talking more about?

@samanthagwolfe @_LucasRizzotto I am trying to get @adrian_kaehler to talk more about it.

When will we have humanoid robots in our homes?

@woputz Sometimes a customer has a very real need that they just don't have the language to articulate clearly.

Any fool can quote. The point is to provide a citation.

@Scobleizer @TZdisruption @Tesla Might be time to get cracking on those virtual beings!

@TZdisruption @Tesla The AI underneath will run: 1. Autonomous vehicles. 2. Robots. 3. Virtual beings. 4. Augmented reality. Same AI. @adrian_kaehler who is a computer vision pioneer has done three of the four and told me they all tie together.

No one is. At this point even Tesla is waiting for Apple. That is why I call it the Big Bang. It will free everyone’s mind.

@mcpli Did anyone notice that one of the markings mentioned specifically in the document is "SI"? Officially "Special Intelligence", this used to be COMINT and generally means signal intercepts. One can only wonder... whose signals?

The front page of today’s NYT

@Scobleizer @elonmusk You will need a hyper scalable low-cost platform that amortizes its R&D cost across all useful tasks.

If AI can drive down the street why can’t a robot wash the dishes @elonmusk and @adrian_kaehler ?

I am absolutely in love with this mouse-over stuff @huggingface API documentations have. It clearly shows they understand their users' pain -- ML APIs have too many arguments. If you are wondering why 🤗 is so popular, little things like these go a long way in reducing friction.

We are on Reddit with a timelapse of our construction shop building one of our robots: reddit.com/r/robotics/com…

How @GiantAITalks uses computer vision in its robot, we call "The Universal Worker." All to help humans with an extra set of hands. youtu.be/QQJwKtk8kPY

@Scobleizer @GiantAITalks Hey, let's not get carried away. I think I might just be a regular smart guy. I'm flattered, really, but Andrej is kind of on another level.

Buy @GiantAITalks. You lost more on Twitter but if you did then you would get @adrian_kaehler and team. He is at Karpathy’s level, built the computer vision system for the first Stanford car.

Now that @karpathy is out of Tesla @elonmusk should get leadership who will aggressively use its data advantages to move into augmenTed reality and robotics. In fact, looking at the team today on LinkedIn, it sure seems that is the case. If not I know how to get both together.

More about the AI/ML/Computer Vision work we have been doing at Giant AI to bring an extra set of hands to workers everywhere: youtu.be/_-awiPpYCpU Our goal is to make it easy to train to do a variety of things humans hate to do (like loading a machine).