There's some truth in this piece. We never quite managed to really have a semantic web because knowledge engineering is actually hard... and we publish mostly unstructured or badly structured data. LLMs are thus used as a brute force attempt at layering some temporary and partial structure on top of otherwise unstructured data. They're not really up to the task of course but it gives us a glimpse into what could have been.
Apparently in the age of people using LLMs for their tests, there is a bias toward mockist tests being produced. It's a good time to remind why you likely don't want them in most cases and limit the use of mocks to consider fakes and checking system state instead.
Interesting stuff about the mathematics behind how embedding spaces work in LLMs.
OK, this is an interesting way for the Darwin Award to branch. Some of the 2025 nominees are indeed funny. Now I wonder which ones will win the award!
We can expect more misleading papers to be published by the big LLM providers. Don't fall in the trap, wait for actually peer reviewed papers from academia. Unsurprisingly the results aren't as good there.
Long but interesting chapter which shows how GPUs architecture works and the differences with TPUs. This is unsurprisingly written in the context of large models training.
Honestly, it took much longer than I expected. Now you know that GitHub has really become a conduit for Microsoft's AI initiatives.
Running interesting models locally gets more and more accessible.
ETH Zurich spearheading an effort for more ethical and cleaner open models. That's good research, looking forward to the results.
Interesting comparison, indeed would a clock like this be useful?
And one more... it's clearly driven by an architecture pattern used by all vendors. They need to get their acts together to change this.
There's always been disinformation in time of wars. The difference is the scale and speed of producing fake images now.
Good reminder that professional translators aren't gone... on the contrary. There's so many things in languages that you can't handle with a machine.
I recognize myself quite a bit in this opinion piece. It does a good job going through most of the ethical and practical reasons why you don't need LLMs to develop and why you likely don't want to.
OK, this is a serious and long paper. It shows quite well how over reliance on ChatGPT during the learning phase on some topics impacts people. It's mesurable both from their behavior and through EEG. Of course, it'd require more such studies with larger groups. Still those early signs are concerning.
Interesting research. Down the line it could help better fine tune models and side step some of the attention system limitations. Of course it comes with its own downsides, more research is necessary.
A nice followup which acts as a TL;DR for the previous piece which was fairly long indeed.
Yep, there's no logic engine buried deep in those chatbots. Thinking otherwise is placing faith in some twisted view about emergence...
An excellent piece which explains well why the current "debate" is rotten to the core. There's no good way to engage with those tools without reinforcing some biases. Once the hype cycle is over we have a chance at proper research on the impacts... unfortunately it's not happening now when it's badly needed.
Or how the workflows are badly designed and we're forcing ourselves to adapt to them.