HNHacker News
TopNewBestAskShowJobs

aabajian

1,743 karma · joined December 29, 2010

aabajian aaaaaat gmail.com 697b79
submissionscomments
aabajian··on The Sphere
If you want a real surreal experience, I suggest visiting Carlsbad Caverns, or a similar enormous cave system. I suspect the awe the author felt was the same sensation as being in an enclosure larger than any building. Think about being in an aircraft hanger then multiple the size by ten. There simply aren't any human-built structures as large. The sky becomes rock and it's just different than any place you've ever been.
aabajian··on U.S. sues Apple, accusing it of maintaining an iPhone monopoly
The blue background on messages sent between two iMessage users has to be one of the most brilliant vendor lock-in strategies. It is an artificial form of discrimination. I feel a slight annoyance whenever a non-Apple user forms a group chat as I know that person will limit the messaging functionality.

In my opinion, the "monopolistic" aspect of it comes down to the fact that they tied it into an otherwise open messaging system - SMS. You cannot separate SMS messages from iMessages (to my knowledge). So, the only way to know a message was sent via SMS is the green background for incoming messages and the green background plus the "sent via SMS" for outgoing messages. This creates a disdain for SMS, and anyone who uses it over iMessage. It is such a strong feeling, that having green messages makes you "uncool", especially in the younger crowd.

On the other hand, I think the long-term sequalae of the blue-green message is to push people to use stand-alone apps like WhatsApp and FB Messenger. I think it'll be a hard sell at this point to convince a jury that iMessage is an overt monopoly.

The main question I want addressed is: If SMS messages can be directly shown in iMessage, and are not secure, then the argument of not allowing "insecure" 3rd-parties to integrate with iMessage goes out the window. All I want is Android messages to be shown in iMessage. Sure we can make them green, but at least they will be sent over the data network and not SMS.

aabajian··on Ask HN: Do you also marvel at the complexity of everyday objects?
I too think about this all the time. Even as a post-doctoral graduate, I still could not make a spoon. Even given a piece of sheet metal, how would I cut, bend and form it into the right shape? And that's a static object made out of one material.

The problem is that you cannot live off of making one spoon a day. Mass-producing everything has diminished the value of one-offs to near nothing. Worse, it has diminished the value of knowing how to make objects. That's why machinists are grossly underpaid.

The closer you get to raw materials, the more price deflation. If you are planning on making something for a living, you'll have to be the most efficient at making a lot of that one object to keep the price down. I met a business owner in Colombia who made "aluminum rods." I don't speak Spanish, and I was confused and asked, "Your company makes rods?" He replied, "Yes, of all shapes and sizes, via extrusion." He is an expert at extrusion. To me, making a rod is harder than making a spoon. Yet, the process is so well understood that there's too many such companies worldwide. There's push by U.S. steelworkers to enact antidumping laws on aluminum extruders from multiple countries (see: https://www.internationaltradeinsights.com/2023/10/petition-...).

aabajian··on Mathematical Introduction to Deep Learning: Methods, Implementations, and Theory
All three authors are PhDs or PhD-candidates in mathematics. The notation is extremely dense. I'm curious who their target audience of "students and scientists" are for this book.
aabajian··on Seventy versus One Hundred
Note that you don't have to assume constant deceleration. You can just assume that the kinetic energy lost by the first car is the same as the second car. I think this is the more reasonable assumption given we know nothing about the breaks or drivers.

The first car loses kinetic energy equal to 1/2 * m * 70^2 while the second car starts with 1/2 * m * 100^2.

Assuming the second car loses the same energy as the first car while breaking, the formula is: 1/2 * m * 100^2 - 1/2 * m * 70^2 = 1/2 * m * v^2 or v = sqrt(100^2 - 70^2) = 71.4 mph

aabajian··on Almost no one pays a 6% real-estate commission except Americans
The MLSes in the US are all under the Realtor group. The selling agent has to agree to pay the buyer agent's fee in order to list. This is one of the main points of the trial. Sellers cannot list a house in the most popular (only?) listing system without committing to the buyer's-agent fee.
aabajian··on After a $1.8B verdict, the clock is ticking on the 6% realtor commission
This was part of the evidence used in the case. They had recordings from buyer's agents who called about non-MLS homes. The agents were on-tape saying things like, "I am going to make sure no other agent shows your house." All because the seller refused to pay a buyer's agent commission.
aabajian··on Redfin Is Leaving the National Association of Realtors
"Most MLSs are owned by the REALTOR® association that formed them." See: https://www.reso.org/blog/mls-faq/

Read the Redfin post. The rules the REALTOR association imposes to get access to MLS data is monopolistic.

aabajian··on Redfin Is Leaving the National Association of Realtors
If Redfin can create a competitor to the monopolistic MLS (owned by Realtor.com), more power to them.
aabajian··on Production AI systems are hard
Drop-down lists (called pick-list in PowerScribe speak) are very common. The issue is that they are completely separate from the image viewer and have no knowledge of the current patient's findings. Admittedly, even a simple classifier could likely choose the "most likely" completion phrase...but again there's a disconnect between the viewer software and the dictation software.
aabajian··on Production AI systems are hard
I'm just finishing interventional radiology training and I moonlight as a diagnostic radiologist (not to mention having an undergrad/master's in computer science).

Almost 90% of the diagnostic studies I read could be pre-drafted by AI. That's where the money is and where AI-in-radiology companies should focus. The money is not in detecting hemorrhage or pulmonary embolism. It's a classic fallacy to think that life-saving means money-saving. Rather, the money in radiology is reads per day.

Here's a user story: A private practice radiologist reads 20 abdomen and pelvis CT scans with contrast per day. In each of these studies, he must write a short description of each organ. For example, "Gall bladder: Unremarkable" or "Gallbladder: Cholecystitis without evidence of cholecystitis" or "Gallbladder: Dependent sludge." There are around 15 such organs (liver, gallbladder, pancreas, spleen, adrenal glands, kidneys, etc.). The AI should auto-populate the radiologist report with an appropriate description of each organ system.

The job of the radiologist is to confirm what the AI says in each section, and to go into further detail as needed. It's essentially just customizing the existing template to each patient. This type of pre-drafting is exactly what radiology residents do and what companies like vRad do.

aabajian··on How to quit cars
Dogs aren't allowed on buses.
aabajian··on Google “We have no moat, and neither does OpenAI”
Recipes!
aabajian··on Google “We have no moat, and neither does OpenAI”
I don't know if I agree with the article. I recall when Google IPO'ed, nobody outside of Google really knew how much traffic they had and how much money they were making. Microsoft was caught off-guard. Compare this to ChatGPT: My friends, parents, grandparents, and coworkers (in the hospital) use ChatGPT. None of these people know how to adapt an open source model to their own use. I bet ChatGPT is vastly ahead in terms of capturing the market, and just hasn't told anyone just how far. Note that they have grown faster in traffic than Instagram and TikTok, and they are used across the demographics spectrum. They released something to the world that astounded the average joe, and that is the train that people will ride.
aabajian··on MiniGPT-4
From a radiology world this is fascinating. I'm not worried about job security as I'm an interventionalist. What I'm wondering is about go-to-market strategies for diagnostics. I do some diagnostic reads and I would love to have something like this pre-draft reports (especially for X-Rays). There are tons of "AI in rads" companies right now, none of which have models that come anywhere close to GPT-4 or even this. Personally, I would not start such a company because the barrier to entry is now so low. I think there needs to be an open-source project that just subtrains these models on rads images and is available to all radiologists.
aabajian··on Pause Giant AI Experiments: An Open Letter
There are Reddit threads showing the jailbreaking of ChatGPT to be more vulgar, and then there are threads showing ChatGPT running its own generated code. Putting 2+2 together, couldn't a jailbroken ChatGPT create and run malicious code?
aabajian··on Hello Dolly: Democratizing the magic of ChatGPT with open models
Anyone care to comment on why the output of these models changes so dramatically given so little Q&A training? It's a 6 billion parameter model with only 50 thousand Q&A samples.

It's clear the model already "knows" the format of a Tweet (short length, attention-grabbing, contains hashtags). The model also knows stuff about language models (word2vec, tokenization), and can include entities from the question in its response (Dolly, Databricks). Yet, it just doesn't put these pieces together in the right way without the Q&A training.

Edit: For kicks, I asked GPT-4 this question: https://imgur.com/a/sM4uyBn

aabajian··on Why buying furniture is so miserable
I came to the same conclusion. The article explains why small-scale U.S. furniture businesses must court designers, and the customers who employ them. It explains essentially what we all know: You can't build furniture at-scale with U.S. wages. I'd be curious to know how a robotic factory for mass-producing furniture would be able to compete with overseas salaries.
aabajian··on GPT-4
I agree that the level of risk/consequence is higher for radiology misses, but I wonder if radiologists are already missing things because of simplification for human feasibility. Things like LI-RADS and BI-RADS are so simple from a computer science perspective. I wouldn't even call them algorithms, just simple checkbox decision making.

This tendency to simplify is everywhere in radiology: When looking for a radial head fracture, we're taught to exam the cortex for discontinuities, look for an elbow joint effusion, evaluate the anterior humeral line, etc. But what if there's some feature (or combination of feature) that is beyond human perception? Maybe the radial ulnar joint space is a millimeter wider than it should be? Maybe soft tissues are just a bit too dense near the elbow? Just how far does the fat pad have to be displaced to indicate an effusion? Probably the best "decision function" is a non-linear combination of all these findings. Oh, but we only have 1 minute to read the radiograph and move on to the next one.

Unfortunately, as someone noted below, advances in medicine are glacially slow. I think change is only going to come in the form of lawsuits. Imagine a future where a patient and her lawyer can get a second-opinion from an online model, "Why did you miss my client's proximal scaphoid fracture? We uploaded her radiographs and GPT-4 found it in 2 seconds." If and when these types of lawsuits occur, malpractice insurances are going to push for radiologists to use AI.

Regarding other tasks performed by radiologists, some radiologists do more than dictate images, but those are generally the minority. The vast majority of radiologists read images for big money without ever meeting the patient or the provider who ordered the study. In the most extreme case, radiologists read studies after the acute intervention has been performed. This happens a lot in IR - we get called about a bleed, review the imaging, take the patient to angiography, and then get paged by diagnostic radiology in the middle of the case.

Orthopedists have already wised-up to the disconnect between radiology reimbursement and the discrepancy in work involved in MR interpretation versus surgery. At least two groups, including the "best orthopedic hospital in the country" employ their own in-house radiologists so that they can capture part of the imaging revenue. If GPT-4 can offer summative reads without feature simplification, and prior to intervention, why not have the IR or orthopedist sign off the GPT-4 report?

aabajian··on GPT-4
I think it will be radiologists signing-off auto-generated reports, with less reimbursement per study. It'll likely result in more work for diagnostic radiologists to maintain their same salary levels.
aabajian··on GPT-4
I'll be finishing my interventional radiology fellowship this year. I remember in 2016 when Geoffrey Hinton said, "We should stop training radiologists now," the radiology community was aghast and in-denial. My undergrad and masters were in computer science, and I felt, "yes, that's about right."

If you were starting a diagnostic radiology residency, including intern year and fellowship, you'd just be finishing now. How can you really think that "computers can't read diagnostic images" if models such as this can describe a VGA connector outfitted with a lighting cable?

aabajian··on Learnings from employing ChatGPT as a ML Engineer for a day
No patient info is sent. In my colleague's example, you can try out a similar query, "Write an appeal letter to a medical insurance company for a patient who needs a biopsy for a bone lesion given prior unclear diagnosis."
aabajian··on Learnings from employing ChatGPT as a ML Engineer for a day
I'm a physician with a software background, another physician at my work used ChatGPT to write a compelling appeal for a patient when insurance denied coverage. I've started to use it to automatically dictate procedural reports. I provide an example report and tell it to modify the report as needed (e.g. "right iliac bone biopsy report with 11G needle"). Beats tabbing (or F2ing) through autotext fields.

These examples make me think, "ChatGPT for form completion" is the killer app as it applies to law, real estate, medicine, etc.

aabajian··on Throughout the rich world, the young are falling out of love with cars
The article starts with a teenager in Portland, OR. If they had selected a teenager just south or north of the city, they would've had a very different perspective.

The value of having and maintaining a car is directly tied to whether you need it to perform the activities you enjoy. I enjoy taking my dog on hikes outside of NYC, so I pay the exorbitant cost of parking ($650 per month on the upper east side) to have a parking spot. If I didn't have a dog, I wouldn't have a car in the city. This is despite the fact that I personally love cars having owned a number of Japanese sports cars.

The focus of the article on climate change activism and urban planning is somewhat political, but I think it gets it right in the end: "Several studies...have concluded that driving habits that are formed in youth seem to persist...even into their 40s."

This is exactly it. I grew up in a relatively remote area of Southern California and became accustomed to driving-in for school, to see friends, to take my dog on hikes, etc. Those experienced have formed a positive association with "owning a car" despite the displeasure of driving around looking for parking.

aabajian··on Building an Internet Scale Meme Search Engine
I had a friend in med school who wrote a very early note-taking app for the iPad. Turns out that there was no way to render PowerPoint files when the iPad first came out. He realized that the iOS/Mac OS "quick preview" function could be used to take screenshots of each PowerPoint slide. For a brief time, his was the only app that could display PowerPoints (albeit, they were just screenshots!). There's a lot of hidden utility in Apple libraries.
aabajian··on Hydrochloric acid is more complicated than you think
Ha, I wonder if that structure provides any more insights into pH.
aabajian··on Hydrochloric acid is more complicated than you think
I remember a bit of this in undergrad gen chem. If pH stands for -log([H+]), how can pure water (H2O) have a pH of 7? In theory, there shouldn't be any hydrogen ions in pure water. Turns out there are two simplifications being made:

i. H+ in aqueous solution is synonymous with hydronium ion (H3O+).

ii. Water is in an equilibrium autoionization reaction with itself and the hydronium and hydroxide ions:

2H2O <=> H3O+ + OH-

A pH of 7 means the negative log of the hydronium ion concentration is 7. Or about 10^-7 hydronium ions per liter. This reflects that the reaction above greatly favors water in the H2O state. We tend to make such simplifications (H+ for H3O+) in order to make teaching easier, but at some point it impedes a deeper understanding of what's really going on.

aabajian··on A fake job offer gone wrong
For everyone who is commenting on the grammar/spelling mistakes, this is on-purpose. It dramatically increases the positive predictive value of their scam (i.e. the probability that the victim they have messaged on day 1 will go through with the wire transfer on day 5). It does this by pre-filtering the general population for gullible individuals.
aabajian··on Ask HN: Has anyone managed to find enjoyment in their work after burnout?
I have some experiencing with the burn-out cycle (you can direct message / email me for my story). Suffice to say, I've done software, teaching (Teach for America), and medicine (I'm finishing residency this year). I have experienced burn-out in every one of these endeavors. What keeps me going is having something else to work on alongside my main grind. These have ranged from side-projects (free apps I made for fun), side-gigs (AirBnb, Turo), side-jobs (part-time programming while in med school), side-hobbies (CrossFit, Yoga, BodyPump, hiking, my dog, dating), and so on. There's no perfect solution, but every once in a while you will find something that you really love and it makes the grind tolerable.
aabajian··on Apple bought out all of Newsweek's 1984 Election issue ad space for Mac
I especially like what is being printed out:

In America, the "Yuppies" -- Young Urban Professionals -- are having babies at an unprecedented rate. And literally nothing is too good for these infants of the affluent. In fact, they're called Gourmet Babies! So we propose to offer traditional Yuppie foods -- sushi, pasta salads, brie cheeses, etc. -- in baby food form.

That business model work work today. Say what you will about Apple, they really know their target demographic.

← PreviousPage 3 of 10Next →