rapatel0

10 ms·

121.
by rapatel0 2y ago
Thanks for pointing this out. I was confused when I initially read the article. Only have access to the abstract but it looks like they are basically converting optical diode to a BJT to modulate bias current. (Might be pulling this out of
122.
by rapatel0 2y ago
Actually, I think it's a fair comparison for several reasons: - The decision of "is it safe to take this trip or not" is tied to the trip. Once you get in a plane or get in a car, you've made the safety decision. It&#x27
123.
by rapatel0 2y ago
Airline ------------ Total number of global flights per year: ~38.9 Million Total deaths over the last decade (3,562) -> 21.2 deaths per million trips Cars ------------ Total number of driving trips per year US: 227 Billion (using your
124.
by rapatel0 2y ago
The writer is 42. As someone approching to that age in a few years, these are standard "mid-life crisis" thoughts. I get them when I feel like shit. Bottomline: - These thoughts don't serve you - You have more effective/
125.
by rapatel0 2y ago
As a bit of a side note, Meta's AI democratization will help with AR/VR. There is a dearth of interesting content to use on them. Open-sourcing will certainly create a fleet of content creators for the platform. I expect the are c
126.
by rapatel0 3y ago
Obviously not, but my general experience with conversations like these is that people never do the math. If we split the profits across all of the people that "deserve reimbursement" you usually end up with a few dollars per perso
127.
by rapatel0 3y ago
you also need to cover the cost of the 100 drugs that failed at various stages of the pipeline. Not saying that it's not a screwed up industry, but stories like these distort the problem as well
128.
by rapatel0 3y ago
Yeah basically. In binary, multiplication is an XNOR operation. 00 = 1 01 = 0 10 = 0 11 = 1
129.
by rapatel0 3y ago
Don't forget Michael https://arxiv.org/abs/1511.00363 < 2015
130.
by rapatel0 3y ago
It's just cryptography and encoding. Popcount is useful for understanding entropy.
131.
by rapatel0 3y ago
The entropy of English is ~1 bit per letter (Measured in a funny way by Shannon - https://cs.stanford.edu/people/eroberts/courses/soco/project... ) In general, it's a bit funny in how we build ML mod
132.
by rapatel0 3y ago
PDF is an incredibly (stupidly) extensible format. There are tons of government forms that (sadly) bake in complex workflows into PDF forms. Given that the whole world has been running on PDFs for decades it's makes more sense to lever
133.
by rapatel0 3y ago
Just to be clear, it's all theoretically possible. There are already versions of BNN versions of YoLo and other CNNs. No reason why transformers wouldn't work for that or audio. It just might be harder to get them to train well en
134.
by rapatel0 3y ago
> 2. United owns several physician networks (totaling 90,000 physicians), who then determine what your care should be—no conflict with the fact that they pay the bills... I disagree that this is necessarily a bad thing. One of the critic
135.
by rapatel0 3y ago
> That’s effectively what much of the rest of the wealthy world does via some form of socialized/nationalized system. While these systems can be more efficient, the rest of the world isn't amazing either. Nationalized systems d
136.
by rapatel0 3y ago
The mathematics of the BNNs are sound. The shannon entropy of a word is really small (I vaguely remember ~2 bits). Also all neural networks are ridiculously over provisioned. I worked on 7 years ago trying to efficiently binarize CNNs from
137.
by rapatel0 3y ago
I’ve never understood The concept of competition in insurance. Fundamentally it’s about pooling funds to average out risk. We had this nonsensical system where we socially subsidize the most expensive (Medicare) and privatize lower risk. It
138.
by rapatel0 3y ago
This is one of the unintended consequences of Obamacare. The act mandates that administration is a fixed percentage of the revenue. Effectively this created the same perverse incentive as cost+ contracts that have hobbled the DoD. There is
139.
by rapatel0 3y ago
Boeing is another in the long list of companies that were taken over by process and finance people and driven into the ground with short term thinking largely centered on reducing cost-structure and financial engineering. Elon has a great d
140.
by rapatel0 3y ago
This is pretty obvious no? LLMs are basically a lossy compression of their dataset. Being lossy, they will necessarily have error (hallucinations). Furthermore, the underlying data is human approximation of truth. Therefore it will have err
141.
by rapatel0 3y ago
Dude should have just put in a hitch and attached flat-bed carrier filled with solar panels.
142.
by rapatel0 3y ago
This article is nonsense clearly written by a compliance person. TLDR “safety good. No safety bad” it fails to articulate any implementation. What problem are you trying to prevent?
143.
by rapatel0 3y ago
Austin is not super conservative. It's actually pretty liberal. Source: I live in austin
144.
AI enabled lives saved from breast cancer – 30 women in 1 year [nature](nature.com)
2 points by rapatel0 3y ago|1 comments
145.
by rapatel0 3y ago
Over about a year, 30 women that were missed by 2 specialist radiologists had their cancers detected, and actioned for treatment with AI. General Article - https://www.ft.com/content/d3b462a3-8085-43d7-9e37-a78564907...
146.
by rapatel0 3y ago
Innovation != using fancy new tech Innovation == using new tech to solve problem and create value.
147.
by rapatel0 3y ago
There are a confluence of reasons behind this state of affairs. - A few fundamental flaws of Obamacare that broke the incentive structure of the healthcare and some regulatory capture. (Granted, the incentive structure has always been a lit
148.
by rapatel0 3y ago
FDA approval is but the first step to actual usage by clinicians and reimbursement by insurance companies. You are probably still 2 years away (maybe more) from being able to ask your doctor to prescribe it and getting insurance to reimbur
149.
by rapatel0 3y ago
That's part of it. The sample size is usually super small. Furthermore, there's the opportunity. With large amounts of data (from software, medical devices, sensors) we can actually tackle this problem at scale.
150.
by rapatel0 3y ago
> Basically, you want to replace statistics ("large enough samples will average out variation") with AI No. This is misunderstanding due to my lack of clarity. Apologizes. The biggest problems with Medical data is that 1) incre

More ›