Loading…

Issue 321: Building good AI ain't so easy

Getting your tech news for the week is as easy as 1-2-3 in Overflow issue 321. What’s not as easy as 1-2-3, though, is building really useful AI. Clearly, there is still a gap between the stuff that works, and the stuff that doesn’t work and defining that gap hasn’t been the easiest for technical folks like us. Sometimes it’s the hardware—that’s probably why chip makers like NVIDIA have entered the model game. On the pod, NVIDIA’s Kari Briski told us about the co-design feedback loop they’ve made between model builders and hardware architects to circumvent some of these issues. Sometimes, it’s how much the model actually knows. On the blog, we’ve got a piece on solving the “context problem” in enterprise AIs. And sometimes, it’s the actual platforms and tools devs work with. Our own Adora Nwodo and Peter O’Connor wrote about how platform teams can become more productive through better organization.

Being able to build stuff that works well isn’t just about AI. The rapidly evolving world of robotics is feeling it too. OpenMind’s Jan Liphardt sat down on the pod to discuss creating open source operating systems for robots that keep Asimov’s Laws in mind. And around the web, people are wondering if building stuff that works is possible during this “tech revolution.” Like, can you even get a robot to build you a PB&J? What about getting an AI model to run locally on your expensive personal hardware? How are you not able to download things outside of the app store when your iPhone is basically a whole computer?

One thing that will always work as easy as 1-2-3, though, is the deep wealth of knowledge coming from our Stack Overflow community. How about we check out some Q&A to close out issue 321. Is it okay to go nude in public if you’re not bothering anyone? Will you finally have control of your life if you have perfect muscle control? Will Gen Z kids call me “unc” if I try to bring back the phrase “Lo!” in conversation? We have all those answers and so much more ready for you in the links below—easy peasy lemon squeezy.

From the blog

Open source for awkward robots

Ryan is joined by Jan Liphardt,  CEO and co-founder of OpenMind, to chat about the rapidly evolving world of humanoid robotics and what it means for humans, why OpenMind is building an open-source operating system for robots that processes logic in natural language, and how putting Asimov’s Laws on the blockchain might be the key to robotics guardrails.

The context problem: Why enterprise AI needs more than foundation models

The difference between AI that impresses people in demos and AI that drives production value is context.

Even the chip makers are making LLMs

Ryan welcomes Kari Briski, NVIDIA’s VP of Generative AI Software for Enterprise, to the show to explore how a chip manufacturer got into the model development game.

Organizing productive platform teams

So many platforms feel heavy because they mirror the organization, not the architecture the organization claims to want.

Interesting questions

Has a Supreme Court recused themselves from deciding a case that affects its own authority?

Is it a conflict of interest to decide whether to check yourself before you wreck yourself?

Would perfect muscle control realistically be advantageous?

Perfect muscle control for who? The person or the implanted chip?

In Massachusetts, why do laws that prohibit nudity require a witness be offended, alarmed, or shocked?

If underwear falls in a forest and no one's offended, is it really indecent?

Links from around the web

Can you instruct a robot to make a PBJ sandwich?

The bots can’t take your job if you work at a Subway.

Willingness to look stupid is a genuine moat in creative work

Willingness to look stupid is also what keeps this Overflow alive.

Your phone is an entire computer

Are you asking if people are down with A-P-P?

Can your machine run AI models?

Your $2,000 custom gaming rig isn’t powerful enough for Llama 3.3, sorry.


Looking for the tools, technologies, and skills your team needs to evolve in the AI era? Stack Overflow's Industry Guide to AI has your answers.