quantumwoke

Wow, what a horrific article. Not the last time that OpenAI has blood on its hands.

I think this does downplay the vulnerability, since it assumes that teams using tailscale SSH have a lax security posture. I would hope that a security conscious team has accounts with least privilege available i.e. no root access. Please don't discount this critical vulnerability.

This might be one of the most consequential S-1s in the next decade. There's a part of me that's terrified, like we're on the precipice of financial or societal collapse. I suppose that could be a sign of the times.

While I agree the job market is tough right now, for context OP is living in Australia in what looks like a regional town, so set expectations accordingly.

I understand what codemode is from other blog posts, and I think this is a suboptimal explanation. It seems that you are explaining it at the wrong level each time, either too deep (here) or too high (elsewhere in-thread when you talk about 'hands' and 'brain').

AI was able to suggest a test for four years and it didn't, and I'm wondering why it didn't.

I think this is where there is simply not enough information in the post. Fibroids can change in size over time. That being said, there are multiple types of fibroids apparently so it may be a type that wouldn't show up on ultrasound.

I got sent this post by a friend yesterday and I read it with my wife (physician). It was interesting although the selective inclusion of details as part of writing a compelling narrative made it difficult to interpret the experience of using AI.

The post seems to portray the fibroid as objectively the cause of the symptoms, although there is no actual evidence of that. My wife tells me that fibroids grow and shrink in size over time, and that larger studies trying to quantify fibroids as a cause of infertility have been inconclusive. A point in time MRI seems to be correlation rather than causation.

It seems the common theme using AI for personal health issues is that post writers (partly due to privacy and length concerns) leave out the gory details like a full timeline of what tests were suggested, what they showed, how many providers were visited, and what treatments were trialled. The timeline is not really clear in this case, but clearly a lot of expensive tests, treatments, and providers happened in the almost four (!) years since they started using AI. Some of those interventions might have happened due to AI advice, and there are clues to this like the mention of possible endometriosis in the biopsy result. Unfortunately, there is just not enough info in the post to know.

That being said, it's clear that AI is helpful for explaining and assisting in advocating for health and luckily there was a positive outcome. Hopefully we can work out the optimal way to incorporate it into the health system.

What was your issue if you don't mind me asking? It's really hard to verify efficacy otherwise.

In this case it took roughly four years for AI to suggest a test. How should we interpret the utility of AI in this situation?

It's so rare in adults it's got the cute name maturity onset diabetes of the young (MODY).

Some observations:

1. It seems at least possible that some of the proof of NS was contained in the training data, making it less novel.

2. The formalisation of mathematics into lean has been an underappreciated force multiplier on discovery.

The named OAI employee has released a statement: https://xcancel.com/SebastienBubeck/status/20973794116915163...

I was six months ahead of vo.codes, well aware. Anyway, UGC was a good pivot and wish you all the best.

You were a year behind 15.ai with FakeYou. I know, because I was six months ahead of you and didn't execute in time. Momentum is everything.

As explained to me, the notes are verbose and do not focus on actionable parts of the consultation. Doctors often have to add a lot of missing detail for legal reasons and if they had to tweak the output through multiple LLMs they would have even less time for their consults.

Not for code, but my wife is a doctor and has gone from using AI scribes back to manually typing out notes and has multiple friends who've done the same. Turns out the gains were eroded by having to check the work and turn verbose prose into an actual note.

Feels like Fable's edge ended up just being long horizon task scaling, which post-training seems to achieve as seen here. Wonder what the next frontier is? Improvement in specialised tasks or computer use?

Just a small meta note: most of the comments in this thread appear to be posting their own codebase (typically AI-generated) that accomplishes the same goal. It's interesting that this problem is simultaneously in high demand and yet considered trivial enough to vibe code per-user solutions to it.

It's his job, and this is advertising (he is very good at native advertising on HN).

And what about the rest of the world? I can't imagine US partners will abide this for long.

This is for the preview period, but it's not a good sign. Opus 4.8 may be the last frontier model available to the masses...

Variations of this comment have been posted for over a year. The pelican has now morphed into part of HN culture rather than a legitimate benchmark, but it's still valuable as a meme.

Exactly my experience. I don't even use em dashes as a signal these days, the sentence structure is enough.

This is terrifying, but I couldn't help myself from frustration at the LLM writing that only worsened over the course of the post. Bloggers, it's not subtle. Please, stop, or at least disclose it.

It's not just transcribing text, there is also a post processing step to turn it from an interview transcript into an actionable summary.

A pretty insightful viewpoint I heard recently from a doctor friend: doctors and hospitals believe that only a corporation could possibly implement this, so they fall into the SaaS trap and lose data sovereignty.

Under the hood, a lot of the companies are Llama or Gemma wrappers connected to whisper.

It's crazy that the hospitals justify this approach. The major scribes are literally connecting Llama to Whisper and that's it.

It's sad that I didn't see any languages on your website from Australia, where there are hundreds of languages that need translating.

There's a lot of people in this thread that don't seem to have caught up with the fact that AMD has worked very hard on their cuda translation layer and for the most part it just works now, you can build cuda projects on amd just fine on modern hardware/software.