Felix Chin MD
@felixchin1
Followers
2K
Following
11K
Media
144
Statuses
4K
aging, genetics, neural nets | Dermatology Resident, MD @Penn, BS Biology @Cornell
Joined April 2012
It hasn’t, @OpenAI has simply tuned GPT-4’s response to match the caliber and quality of the prompt. GPT-4 will only give good responses if the prompt shows deep understanding and thinking. This is how they’re making it a personalized tool for all. Currently putting together a.
This is huge. Recent academic paper shows that GPT-4 lost significant capability in code skills over a few months. I love GPT-4 but it would be nice to have OpenAI address this. They've been insisting that users are "imagining it," basically.
205
126
1K
@stevesi Conspiracy theory: Apple is preventing other companies from collecting data so they secretly will have an edge in data-curation moving forward. All under the public perception that they’re the benefactors of “consumer privacy”.
4
8
261
@emollick Prompt engineering is the future. The ability to explain a problem well and in plain language will become invaluable.
16
12
153
@netflix Really hoping that there will be a few episodes dedicated on AI and worst possible outcomes in terms of a dystopic AI society. Would make it extremely relevant to current discussions that are being had on AI safety and responsible implementation.
4
1
151
1
3
148
@Chrisprucha @OpenAI The issue is when GPT-4 gives a response that’s deeper or smarter than what the user is able to comprehend, the user will complain, or get uncomfortable. Think about real life interactions, when you meet someone in real life who suddenly tries to have a deep conversation with.
51
11
122
.@sama gets 7%, “this is wrong, openai should be a non-profit”. sama doesn’t get equity, “he must have an ulterior motive, we can’t trust him”. as jobs once put it, “u can only make some of the ppl happy, some of the time”. unpopular opinion- i actually think altman’s genuine and.
this is very much it: investors are driving the demand for altman to take a stake
13
8
124
@elonmusk @DrKnowItAll16 You should moderate a discussion between Sam and Ilya on 𝕏 spaces right now.
4
2
86
@chrisprucha personal compute clusters (dozen m4 mac minis) are what’ll drive future discoveries . greatest breakthroughs originate from an individual, not 100 ppl sharing a giant centralized data center.
6
3
85
@wongmjane Seems like there's a hidden race and battle between these big tech companies when it comes to fighting each other for user data. Looks like these companies are all trying to one-up each other to build their software architectures so that they'll have an advantage over the other.
2
2
68
@ParikPatelCFA @ParikPatelCFA is a proud papa.of his 3 sons: @chamath , @VivekGRamaswamy , @Kash_Patel . but still disappointed that none of them are doctors like him.
4
1
67
@chrisprucha the tests used here (blender) don’t even use mlx acceleration. with mlx, matmuls are likely an order of magnitude faster than the 4080 (pytorch). m4 max is likely way faster than 4080 for developing/running neural nets (if u use mlx for matmuls and backprop).
2
1
59
@vkhosla English so happens to be the language by which we interact with AI. In other words, it's a very powerful programming language now that large language models are on the rise. Could be useful to be an English major, though this probably wasn’t the reason why most English students.
4
1
57
@karpathy such a system would not be completely deterministic tho which is one of the key reasons why computers are so useful.
4
0
58
@VPrasadMDMPH It's a byproduct of our flawed education system. Universities rarely teach critical thinking and conceptual understanding anymore. This goes double true for medical schools, which is essentially a marathon of rote memorization and pattern recognition. Science comes from actually.
4
7
58
@paulg @charliekirk11 the more ‘1940’ comparisons u make, the more it diminishes the significance of the events that occured that decade.
1
0
56
People give me strange looks when I tell them how interested I am in artificial intelligence (AI), data science and physics as a medical student. The reason why I’m so interested in these disciplines is because I want to apply them onto medicine. I think there’s an enormous
6
4
53
@paulg programming is powerful because it’s deterministic. entering prompts into an AI is not deterministic. therefore using AI to write code will never be the same as writing code oneself.
17
2
50
if insurance covered everything doctors requested, our country would go bankrupt immediately.
"A lot of health care that doctors want to give, but insurers don’t want to cover is 100%, complete and total bullshit.".
2
3
53
@Quikolas17 @OpenAI Good question = solid understanding of the problem that is trying to be solved.
2
1
50
@karpathy ai detects the ‘vibes’ of a prompt.and returns a response of similar ‘vibes’ that best overlaps with the collective human language it was pre-trained on. it’s a vibe machine.helps the user discover (generalize to) information on the same vibe-frequency as them.
5
3
48
@typedfemale The mind is literally like a muscle. If you don't use it, you lose it. It's no different than maintaining physical muscle mass through regularly going to the gym. If you stop going to the gym, you'll flab away.
0
3
45
Another reason why OpenAI getting acquired by Microsoft is a good thing. Google owns the patent for the transformer model. If it weren't for Microsoft taking OpenAI under their wing, Google could sue OpenAI into oblivion and give GPT-4 the axe. In other words, Microsoft is like
6
4
43
@lexfridman Don’t forget courage to report, even when it comes at the expense of their own interests. I think that’s the most difficult quality to come by given the personal sacrifice required.
1
0
44
@pmarca signal density so high gotta 0.5x it. vs college lectures which are filled w so much fluff u gotta 5x it just to stay awake.
1
0
44
@MrBeast When it comes to being jacked, a lot of it also comes down to genetics. So don't feel too bad if you’re unable to achieve your dream physique. Most important thing is to be the best version that you can be and find pride in that version of yourself.
10
1
37
@then_there_was the original openai team were researchers who enjoyed discovering things. now that they ‘know what to do’, the culture must be shifting from discovering to executing. the ppl leaving are probably looking for the next big thing to discover.
3
1
40
@jasonryanmd This makes too much sense for a culture that no longer appreciates common sense.
0
0
35
Why WiFi waves aren’t harmful to the human body:. In terms of the electromagnetic spectrum, Wi-Fi waves oscillate at a frequency of 5 x 10^9 Hz. When you compare this to the least powerful visible light (red light, which we’re surrounded by constantly) which operates on a.
Believing wifi opens the blood brain barrier and causes toxins to enter is crazy, imo, but RFK Jr is not smashing the router in my home. Meanwhile, public health 'experts' who think themselves too brilliant to debate made toddlers mask for years-- equally fucking crazy.
12
3
39
@yacineMTB This is coming from someone in medicine and biomedical research where peer review takes months, if not sometimes years. Am I correct in assuming that the peer review process for computer science and machine learning research mostly takes place on Twitter, where the best papers.
5
2
36
2
2
37
@Tesla_AI We talk about OpenAI, Google and Meta being the leading forces in AI. I'm willing to wager that Tesla has all of them beat with their AI talent and engineering team. Seems like they're not only designing AI models but they're also integrating it with hardware. Similar to how.
3
1
34
@lindayacc Twitter should build a dating app into its platform. It's much easier to gauge a person’s personality and interests by looking at their Tweet history than it is by glancing at a superficial dating profile, which is usually contrived, artificial, and unreflective of what their.
14
1
33
@unusual_whales Doesn't that equate to an annual inflation rate of 2.6% per year? I don't understand how this is anything shocking or out of the ordinary unless you're trying to test people's basic understanding of whether they know how annual inflation works.
4
0
30
@FutureJurvetson uh… what is mythic m2 and is it rly 2 orders of magnitude more powerful than elon’s (xai) colossus?.
3
4
32
@ProfFeynman Teaching well requires clarity in thought. Clarity is a sign of deep understanding of a topic.
0
3
30
@naval would love it if u could provide ur interpretation (summarization) of the book in ur own words. my favorite part of the deutsch files isn’t when david deutsch speaks, but the tiny moments when u provide a compressed summary of what he says . a video of just ur.
5
0
30
@OpenAI Looking forward to this increase. GPT-4 has been an incredibly powerful tool in my research towards studying individual molecular differences between lupus patients (a disease for which we currently lack effective treatments for).
3
0
30
@pmarca the donation will help establish a free speech committee which’ll create rules/regulations around what u can/cannot say (in the name of free speech) which paradoxically will decrease the amount of actual free speech on campus.
3
3
30
@ML_PhDer And also a great reason to give these undergrads to more seriously consider industry over academics.
0
0
30
@unusual_whales Always a big difference between academic theory and what actually happens in real world practice. The real world is much harder to predict than in academics where many factors are nicely controlled for or ignored and disregarded.
1
0
30
@philosophors Those same people would be extraordinary if they could just be themselves. The pressure to conform is great these days.
2
0
28
@ElonMuskAOC Sometimes I wonder if this parody account is actually Elon Musk beta testing different tweets to see what the public's response might be.
2
5
26
@mckaywrigley Whether or not this is useful for solving any real problem is a question worth having.
3
0
28
@kareem_carr Discovering something yourself is far more impactful than just having it taught to you. In some ways this woman did invent calculus. Just because it already existed in the world doesn't mean that it wasn’t an amazing feat of creativity and ingenuity on her part. Also,.
0
1
21
@alifarhat79 It's only illegal when people know about it. The unfortunate reality is that there are a lot of hidden companies that are very good at hiding and being very discreet about the fact that they have complete ownership and control over various sectors of the economy.
1
1
22
OpenAI should make their own version of Siri. They already have Whisper for voice recognition. ChatGPT can generate the response. The Speechki plugin can read ChatGPT's response out loud, serving as Siri’s voice. What do you think? @sama.
14
0
24
@calleymeans @DrNeilStone it’s concerning how nonchalantly ppl are pulling the holocaust card these days.
1
1
23
My hypothesis is that GPT-4 is a reflection of the user. The deeper and more well-thought out the prompt, the deeper and more personalized the response. Also very important to provide context behind your question. If you feed GPT-4 a generic templated prompt, expect a generic,.
The amazing Felix Chin surfaces how most AI models will require SuperPrompts to be powerful. The complexity of the prompt will determine the complexity of the outputs. Why a will always be needed.
7
1
19
@DrJimFan when trying to build something. the bottleneck is reality, everything else is a dream. put 100% of ur attention on fixing the bottleneck. once the bottleneck is fixed, move on to the next bottleneck. solve bottlenecks one after the other, in rapid succession, and ur vision.
0
0
22
@nikitabier The need to raise money is a reflection of one’s dependence on others to do the actual work.
3
1
22
@Andrew_Akbashev if nobel prize was about number pubs or h-index then medical students applying into dermatology would be top contenders.
1
0
22
The concern with academia these days is that people are more concerned about getting published than actually discovering new and useful knowledge.
3
3
19
@stats_feed It honestly boggles my mind when I hear so many of my medical school classmates say that New York City and San Francisco are their dream places to live and practice medicine in as future physicians. Not only is the cost of living much higher in those cities, but physician pay is.
1
0
20
@Yampeleg The only people who use Auto-GPT are the programmers who are making it and then finding it to be useless when it comes to any practical real-world benefit.
2
0
20
@RampCapitalLLC For people who don't like caffeinated beverages to enjoy the fine taste that coffee has to offer. Yes, I am one of these weird individuals.
2
0
18
@Jason vertically integrated approach to age reversal (epigenetic reprogramming). step 1, discovering precise protein targets using genome-wide crispr screens. step 2, AI drug design to develop molecular glue “binder”. step 3, clinical trials in wound/ulcer patients to demonstrate.
1
0
10
@foundertribune @justinemusk her 3rd ted talk on visionaries and dark creativity was one of the most inspiring things ever (esp the ending). has significantly shaped the way i approach life.
1
4
19
@ML_PhDer The transformer model was published by machine learning scientists at Google. We tend to think of Microsoft/OpenAI and Google as competitors but to some degree it’s also a community effort.
1
1
18
@BrianJJi @SpiceBaked MBAs are a nice status symbol. Though it's hard to get anything groundbreaking/original from people who place great importance on social validation, which is essentially a prerequisite for a MBA, and for fitting in.
0
1
19
@yacineMTB I could train an AI to automate her expertise and then deploy 10 thousand of her in simulation and show her work results in nothing useful even when done at scale.
3
0
20
I definitely agree GPT-4 has been nerfed when fed shorter prompts. However if you provide it with lots of detail and most importantly, context to your question, I’ve found that the responses have never been better.
As someone who uses GPT-4 every day, I feel that ChatGPT seems lobotomized. It's a real problem.
6
1
16
@CastorT26371809 Yes, that's true, but microwave ovens typically operate at a power of 1000 watts. Contrast this with Wi-Fi routers which operate on less than a single watt. The waves from Wi-Fi contain far less energy than that from microwave ovens. Good question though. Wish more scientists.
3
1
19
@pressf1 @appltrack @verge This seems like the type of comment that should become a community note. @CommunityNotes.
1
0
17
@garrytan @alexandr_wang I used to be miserable back when I tried to be normal. Could never succeed in becoming “normal”. I’m glad I didn’t, otherwise I wouldn’t be doing what I am today. The hardest part is realizing that “normal” isn’t actually that cool. Once you come to that realization, it becomes.
0
0
15
@jasonryanmd dont forget all the trivia that we’re made to memorize that not even cardiologists know/remember.
1
0
18
@davidasinclair It honestly blows my mind when people can't seem to understand that aging is the fundamental driver behind so many of the diseases that plague our society. And is also what ultimately causes death.
1
1
16
@lexfridman Chances are they’re all single-planet civilizations, confined to their own dustballs until they’re annihilated by an extinction level event. This is why it’s important we reach Mars.
3
0
16
@mcuban @elonmusk @TruthGpt @Twitter My god. The idea of training GPT off the entirety of twitter is genius. And it’d allow @elonmusk to avoid any potential legal ramifications, unlike OpenAI which is currently getting sued from every direction from a wide variety of platforms having used their data.
4
1
14
@OptimismMommy same phenomenon exists in medical-education (med school level and above). nicest ppl in the world, with semi-decent theories. what they lack is the courage to execute (willingness to be disliked) and the ability to error-correct when they’re wrong.
1
1
17
@__nmca__ allowing pdf attachments would make it dramatically more useful for scientific research.
1
1
17
@genioprompts @OpenAI Less about engineering and more about just being able to explain a problem well in plain language. Also demonstrating a deep understanding of the problem you’re trying to solve.
1
0
14
@davidasinclair Is there much single-cell research being done in aging research? Would it be useful? Curious because this is where I’m trying to build expertise (especially in computational analysis and AI-driven discovery) and I’m looking to dedicate a career towards age-reversal science.
3
1
16
@paulg @USTechWorkers Talented immigrants are a huge part of what drives this country and makes the US so successful.
1
1
16