phren0logy

7 ms·

61.
by phren0logy 1y ago
I don't know your tools, but if you aren't recommending them to others for money, then we're already in a different category. But succinctly, if something can help you form habits, it can probably help you form bad habits.
62.
by phren0logy 1y ago
I was not familiar with it until now. At a quick glance, it looks like it has some overlap with the CPT. But generally speaking, this is going to give a snapshot in time in one setting, which is helpful, but just one part of the picture. As
63.
by phren0logy 1y ago
I'm a psychiatrist, so I obviously have a bias, but I have been pretty alarmed by the caviler attitude to the diagnosis and treatment of ADHD and Autism-spectrum disorders. This is not specifically directed at this product launch, rath
64.
by phren0logy 1y ago
I have not heard this rumor. Source?
65.
by phren0logy 1y ago
This worries me because I think LinkedIn has a set of assumptions and usage patterns that I would not want to see spread... well, anywhere, really. But certainly not to Office. It could be this person is savvy enough to avoid that problem,
66.
Microsoft's LinkedIn chief is now running Office as part of an AI reorg(theverge.com)
3 points by phren0logy 1y ago|2 comments
67.
by phren0logy 1y ago
Is there a link to an unpaywalled version? I always appreciate Thompson's insights, but I don't read them for investing advice, so the value is hard to justify for my idle curiosity alone.
68.
by phren0logy 1y ago
My use case is for documents related to a legal issue, where a foundation model has no knowledge of any of the participants or particular issues. There are many, many such situations. Your statement is ignorant and overly broad.
69.
by phren0logy 1y ago
An AI app for trauma survivors sounds superficially laudable, but I really hope they are working with seasoned professionals to avoid making things worse. Human therapists can sometimes make things worse, too, but probably not at the same s
70.
by phren0logy 1y ago
It's called a "waitlist" control group, and it's not intended to represent placebo. Or at least, it shouldn't be billed that way. It's not an ideal study design, but it's common enough that you could use i
71.
by phren0logy 1y ago
If the idea of "spaghetti western theremin" strikes you as at all interesting, check out Via Mardot! https://youtu.be/jfpYluPyP3s?si=ohiXEkdEYoJkFht0
72.
by phren0logy 1y ago
> in the drug treatment world it’s basically known that drug abuse is almost always a reaction to some kind of psychological trauma or disorder This is incorrect. While this is true for a substantial number of people, I want to offer som
73.
by phren0logy 2y ago
That is how I read it. Transformer based LLMs have limitations that are fundamental to the technology. It does not seem crazy to me that a guy involved in research at his level would say that they are a stepping stone to something better. W
74.
by phren0logy 2y ago
I guess I'm on the other side of this. $30-100 is a reasonable price for books for a college-level course in the United States. Assuming this https://educationdata.org/average-cost-of-college is accurate, tuition at th
75.
by phren0logy 2y ago
This looks very cool, but I'm too chicken to be first in line to try it. I will, however, be keeping an eye on it!
76.
by phren0logy 2y ago
I'm a bit surprised at the approach to RAG. It will be great to see how well it handles complex PDFs. The max size is far larger than the Anthropic API permits (though that's obviously very different - no RAG). The chunking stra
77.
by phren0logy 2y ago
That's a real issue, but that's masking some of the issues further downstream, like chunking and other context-related problems. There are some clever proposals to make this work, including some of the stuff from Anthropic and Jin
78.
by phren0logy 2y ago
I think this is a fundamental issue with naive RAG implementations: they aren't accurate enough for pretty much anything
79.
by phren0logy 2y ago
FYI, you can choose which OCR engine Docling uses (from a handful of predefined choices) - it doesn’t have to be Tesseract. https://ds4sd.github.io/docling/reference/pipeline_options/#...
80.
by phren0logy 2y ago
I should have mentioned that also: getting access to HIPAA-compliant models was the tipping point for me.
81.
by phren0logy 2y ago
I am guilty of this, to a degree. I tried every local model around the time of the original llama 3 release, and for a while after that. Mostly the models around 70b. They always felt like toys compared to the SOTA hosted foundation models.
82.
by phren0logy 2y ago
For me, it's also about voting with my wallet. I'm not enthusiastic about invasive ad tech. As it stands, nobody else offers what Kagi offers at any price. If there were an equivalent service for $5/month, I'd give it a
83.
by phren0logy 2y ago
I think people aren't getting your joke.
84.
by phren0logy 2y ago
Interestingly, in some informal experimenting I have found these issues are somewhere between "much reduced" and "totally absent" when running the model locally. This appears to be an additional layer of guardrails appli
85.
by phren0logy 2y ago
You forgot Slow Horses!
86.
by phren0logy 2y ago
I'm sure it will improve over time. I'm just surprised it was as bad as it was at launch, and has improved minimally.
87.
by phren0logy 2y ago
Given how hard Microsoft is leaning into AI, and how important Office 365 is, the Copilot for 365 is shockingly bad.
88.
by phren0logy 2y ago
If this scenario is relevant for anyone, it's a tiny number of people. That leaves 99.99% of the fitness market.
89.
by phren0logy 2y ago
In my use, docling has not involved an LLM. There are a few choices for OCR, but I don't think a vision model is one of them. It's certainly touted as a solution to digest documents into plain text for LLM use, but (unless I just
90.
by phren0logy 2y ago
It's easy to turn out 10x more mediocre ad copy. It's a lot different to actually learn and comprehend 10x more material. These are not even remotely equivalent tasks.

More ›