HNHacker News
TopNewBestAskShowJobs

johntiger1

939 karma · joined October 7, 2017

radiantai.health/
submissionscomments
johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
We've built an automated scraper and data ingestion ETL, which crawls search engines for medical PDFs. We'll definitely need to add in compliance and verification to make sure we're not crawling any protected documents.
johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
Yep, as a medical student, I strongly believe we need explainable AI for healthcare, and that AI only empowers clinicians to make the final call, and does not do any medical decision making autonomously. I really hope other players in the space like Hippocratic AI [1] and Google Med-PaLM2 [2] understand this as well.

[1]: "www.reuters.com/business/healthcare-pharmaceuticals/hippocratic-health-raises-50-mln-seed-funding-build-ai-model-2023-05-16/"

[2]: https://cloud.google.com/blog/topics/healthcare-life-science...

johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
Yes, it's retrieval augmented generation, aka IR + NLP
johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
Thanks for all the queries. We're still very much WIP - didn't even mean to launch til next week and suddenly found our site on HN! Will be adding a new module which should address all of these questions :)

But one reason that you are getting these is that our underlying knowledge base lacks specific knowledge on these topics. However, that's only a question of time as we scrape more documents from the web.

johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
Right now, we've found OpenAI to be far superior to any of the other offerings (Anthropic, Meta) [1]. As much as we want to support a fellow Canadian startup (Cohere - which actually gave us free credits - S/O to them), we found their generation quality severely lacking.

[1]: https://hermitian.substack.com/p/claude-and-llama-get-basic-...

johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
Yep, that's exactly why we're building it. Our goal is to make generative output explainable and interpretable. Right now we're doing this by showing the retrieved context, which is fed directly to the LLM to perform generation. You can then verify for yourself whether or not the generated content is hallucinated from the retrived document
johntiger1··on Show HN: ChatGPT for Med-School and Healthcare
Wow, didn't expect to see my own startup launched on hackernews like this :) But thanks for all the feedback so far! Will be taking the highlights back to the team
johntiger1··on Show HN: Summary Cat, a YouTube Video Summary Generator
Why though? From a user perspective, it should return a 5xx, not hang endlessly
johntiger1··on Show HN: Summary Cat, a YouTube Video Summary Generator
I am thinking about music videos, where the lyrics don't describe the video necessarily. For instance, I am curious about what exactly is the story in this music video: https://www.youtube.com/watch?v=pruKV1chnHA&ab_channel
johntiger1··on Show HN: Summary Cat, a YouTube Video Summary Generator
Interesting, very cool!

However, how does it do on videos where there's not a lot of speaking? Any plans to do actual video (image) processing?

johntiger1··on An Inside Look at TikTok from a 3 year employee
Informative, but please refrain from posting Blind links
johntiger1··on Have you received AI generated spam pull requests on your open source repo?
Not on my research github repo for my paper, no.

[1]: https://github.com/johntiger1/multimodal_fairness

johntiger1··on Ask HN: 6 months later. How is Bard doing?
Can you share more about how it was 'spooky'? Like it was completely unregulated?
johntiger1··on Ask HN: 6 months later. How is Bard doing?
I believe it is region locked. So people in Canada etc. only got it recently
johntiger1··on Subdomain.center – discover all subdomains for a domain
Took a while, but was impressed it detected all of ours: https://api.subdomain.center/?domain=radiantai.health
johntiger1··on Claude and LLaMa get basic human anatomy wrong
It's literally being taught in classes at a major Canadian medical school, in 2023.
johntiger1··on Claude and LLaMa get basic human anatomy wrong
Not in Quebec or the UK.
johntiger1··on Claude and LLaMa get basic human anatomy wrong
I was testing on GPT3.5-Turbo, the free tier.
johntiger1··on Claude and LLaMa get basic human anatomy wrong
Don't think I learned about monoids in 1st year CS. I have learned "I ate 10 eggs at 12" in 1st year medical school.
johntiger1··on Claude and LLaMa get basic human anatomy wrong
Where are you getting the 1.7T number from? Genuinely curious, as sources like [1] generally cite 175B.

[1]: https://ecoagi.ai/articles/chatgpt-parameters

johntiger1··on Claude and LLaMa get basic human anatomy wrong
Interestingly, we actually do exactly that, although our underlying LLM is OpenAI-based, so we would have avoided the error in the first place :)

> I disagree with the idea that this is basic human anatomy

When I say basic, I mean something taught in second or third year anatomy or physiology courses; or first year medical school.

johntiger1··on Claude and LLaMa get basic human anatomy wrong
I don't think it's that obscure. It's something taught in first-year medical school, or in undergrad anatomy or physiology. Hardly anything that deep
johntiger1··on Claude and LLaMa get basic human anatomy wrong
Interesting, haven't heard of Pi. Is this a custom LLM?
johntiger1··on Claude and LLaMa get basic human anatomy wrong
Perhaps more time with RLHF due to first mover advantage? If you figure it out, Meta and Anthropic would probably pay big $$ for the secret
johntiger1··on The Tyranny of the Marginal User
The internet really took a turn for the worse once they let people like me on to it.

/s

johntiger1··on ChatGPT web and mobile UIs unavailable
I'm using claude while it's down, and it's a fine alternative
johntiger1··on How to not get rejected from YC's Early AI interview batch
Yep, glad you got the point. YC isn't some test where "passing" is the goal. As trite as it sounds, you are evaluating them as an accelerator as much as they are evaluating you as a startup - if the fit isn't there, either the startup changes itself or you move on. No big deal either way.
johntiger1··on Tesla created secret team to suppress thousands of driving range complaints
Does all this somehow not apply to ICE vehicles? Cause the gas meter/range estimate was pretty accurate in my dad's '93 Corolla...
johntiger1··on Tesla created secret team to suppress thousands of driving range complaints
It's a very humbling experience to be ignored by so many folks, glad you reflected on that. The story about the couple just flat out walking away strikes me as obtusely anti-social, but sadly not unexpected in our day and age.

Just out of curiosity, where was this?

johntiger1··on Autoenshittification. How the computer killed capitalism. – by Cory Doctorow
I think Cory is going a little too harsh on Uber/Amazon. People use them because they provide convenience and ease-of-use, and are willing to pay for a better customer experience. If anything, 'taxi medallions' are literally the original form of feudalism and collecting rent from doing nothing.
← PreviousPage 3 of 9Next →