Iroh 1.0 - Dial Keys, not IPs

Iroh 1.0 is out. Now is the time to build.

No, Artificial Intelligence Is Not Conscious

Taken to its logical conclusion, this line of thinking is absurd—and damning.

Leo’s first encyclical attacks technological messianism

The Pope warns against replacing humans with AI, though he seems to use it himself

The Death of the Brick & Mortar Toy Store

It doesn’t take a genius to figure out why more and more local stores are going defunct. A …

Have a Coherent AI Policy

Yet Another Software Engineer's Blog

People Hate AI Art

Hallucination is Inevitable: An Innate Limitation of Large Language Models

Hallucination has been widely recognized to be a significant drawback for large language models (LLMs). There have been many works that attempt to reduce the extent of hallucination. These efforts have mostly been empirical so far, which cannot answer the fundamental question whether it can be completely eliminated. In this paper, we formalize the problem and show that it is impossible to eliminate hallucination in LLMs. Specifically, we define a formal world where hallucination is defined as inconsistencies between a computable LLM and a computable ground truth function. By employing results from learning theory, we show that LLMs cannot learn all the computable functions and will therefore inevitably hallucinate if used as general problem solvers. Since the formal world is a part of the real world which is much more complicated, hallucinations are also inevitable for real world LLMs. Furthermore, for real world LLMs constrained by provable time complexity, we describe the hallucination-prone tasks and empirically validate our claims. Finally, using the formal world framework, we discuss the possible mechanisms and efficacies of existing hallucination mitigators as well as the practical implications on the safe deployment of LLMs.

Repairing the Ruins: Why AI Can’t Replace Education

COMMENTARY: In an era when AI can write anything, authentic education must go beyond the mere production of words.

Sneaky spam in conversational replies to blog posts

I'm grateful that my blog posts attract lots of engaged, funny, and challenging comments. But any popular post also attracts spammers. I use Antispam Bee to automatically eradicate a couple of hundred crappy comments per day. Nevertheless, some get through. Here's a particularly pernicious one - it appeared as three comments ostensibly in reply to each other. At first glance these look like …

What we learned building 200+ API integrations with OpenCode | Nango Blog

We built a background agent for API integrations. Here is what worked, what failed, and how we made it reliable.