It's a radical change in human/computer interface. Now, for many applications, it is much better to present the user with a simple chat window and allow them to type natural language into it, rather than ask them to learn a complex UI. I want to be able to say "Delete all the screenshots on my Desktop", instead of going into a terminal and typing "rm ~/Desktop/*.png".
I feel like we are due for a revolution in computer interface design that will free us from our desks. I want to be able to do work while walking on the bike trail or sitting in a lounge chair by the pool. All the core concepts of GUI design - "mouse", "window", "file", "folder", "desktop" - were developed in a previous era with far tighter constraints on what could be done. Now we have voice understanding, wearable computers, AR / VR, LLMs, cellular internet, etc. Even though the tech has advanced by leaps and bounds, the underlying UI concepts haven't changed much.
The key to the power of the LLM is that the training process can learn effectively from vast corpora of unlabelled text. Unfortunately, there is no comparably vast database of medical images.
In order to "crack" radiology, the AI companies would need to launch an enormous data collection program involving thousands of hospitals across the world. Every time you got an MRI or X-Ray, you would sign some disclosure form that allowed your images to be anonymously submitted to the central data repository. This kind of project is very easy to describe, but very difficult to execute.
There is a classic capitalist ploy, documented in Steinbeck's Grapes of Wrath, where business owners vastly over-advertise the abundance and remuneration offered by a certain area of work. Then many hopeful people uproot themselves to go into that field, so supply and demand pushes down wages, benefitting the owners.
Note that you can make things even simpler than this: just use vanilla JS to compose HTML strings, and plug them into divs using innerHTML = "...". When your state changes due to user input, just rebuild the page using the new data.
I've been building many simple apps using this technique, it works fine. I'm sure there is some kind of performance hit; it's almost never an issue. I have never seen any browser compatibility issues.
This is one of a whole genre of HN-boosted blog posts that are basically saying to engineers: "you can't just write code, you actually have to do sales also". People who don't like sales think that they can go to a big company and just do engineering work. Nope, it doesn't work that way: you still have to promote your work, you still have to sell your work, you still have to talk to your customers, you still have to understand business, you still have to make deals, etc etc.
I love Vanguard, and I would like see more companies like them. They offer a great service to millions of people, manage vast amounts of money, but the top leaders are not multi-billionaires.
It's weird that the mainstream TV execs think audiences want boring American accents. To me, one of the best things about the White Lotus (hit HBO show) is that it highlights a distinct array of accents (including Australian).
For me, the futility of GOFAI is obvious from simple introspection: is my own brain capable of chaining through vast networks of propositional statements? I wish! I can barely hold a 5 node decision tree in my head while playing a board game. At the same time, my brain is amazingly good at parsing visual and linguistic stimuli.
Still, you have to respect the man's dedication to his quixotic quest.
One of my pet theories about the software industry is that nobody really knows how to manage mature tech companies (YC and the startup world are pretty good at running young companies).
One obvious and disastrous phenomenon in the tech world is resume-driven development: some engineers are highly motivated to put the next shiny tech buzzword on their resume, so they make sure to push that technology at their company. 9 times out of 10 the project and company would be better off by just using the standard, boring tech that everyone else uses. Tech managers should be able to detect this pattern and squash it, but they don't seem able to do so.
For questions like this, you need to tell the bot to search the web.
> Hi, can you search the web and find out if Paul Newman has any alcohol problems?
> Paul Newman, the acclaimed actor and philanthropist, had a well-documented fondness for alcohol, particularly beer. He was known to consume significant quantities, with some accounts noting he would drink a case of beer in a day. This level of consumption led some to describe him as a functioning alcoholic. His drinking habits reportedly caused strain in his marriage to Joanne Woodward. In the late 1970s, an incident occurred where Woodward did not allow Newman back into their house because of his drinking. He promised to give up hard liquor as a compromise, but continued to drink beer, sometimes consuming a case during dinner.
You should train a GPT on the raw data, and then figure out how to reuse the DNN for various other tasks you're interested in (e.g. one-shot learning, fine-tuning, etc). This data setting is exactly the situation that people faced in the NLP world before GPT. I would guess that some people from the frontier labs would be willing to help you, I doubt even your large dataset would cost very much for their massive GPU fleets to handle.
I view the current political situation as a historical rhyme of the bloody wars of religion in Europe that saw Catholics and Protestants murdering each by the millions. As with the C v P wars (or Sunni vs Shia in Islam, or Hindu vs Muslim in India), taking a side is the wrong ethical stance; the right stance is to view it clearly as a brutal tribal struggle, and take steps towards a just and honest peace.
I might be the only person in this camp, but I find the "standard" command line arguments style absolutely repulsive. I write tons of CL code, and I always use easy key=value notation (sometimes it's flag=true, which I consider to be a minor sacrifice of conciseness in favor of consistency and readability).
DNNs do not have special generalization powers. If anything, their generalization is likely weaker than more mathematically principled techniques like the SVM.
If you try to train a DNN to solve a classical ML problem like the "Wine Quality" dataset from the UCI Machine Learning repo [0], you will get abysmal results and overfitting.
The "magic" of LLMs comes from the training paradigm. Because the optimization is word prediction, you effectively have a data sample size equal to the number of words in the corpus - an inconceivably vast number. Because you are training against a vast dataset, you can use a proportionally immense model (e.g. 400B parameters) without overfitting. This vast (but justified) model complexity is what creates the amazing abilities of GPT/etc.
What wasn't obvious 10 years ago was the principle of "reusability" - the idea that the vastly complex model you trained using the LLM paradigm would have any practical value. Why is it useful to build an immensely sophisticated word prediction machine, who cares about predicting words? The reason is that all those concepts you learned from word-prediction can be reused for related NLP tasks.
I don't know about this project specifically, but this kind of work is soon going to be tremendously valuable because of AI. Instead of engineers designing products directly, they will build a simulator, write an objective function, and submit this data to a general purpose ML/RL API hosted by one of the big labs. The AI will run a billion simulations and use RL to create a design that optimizes the objective function.
I'm terrified some colleague is going to vibe-code a product and use it to get promoted, then move to another team and dump the project on me to maintain.
Blogging is like all other content creation: the vast majority of the viewership goes to the top 0.1% of creators. This is the same whether you're talking about YouTube, Twitter, podcasts, OnlyFans, scientific papers, traditional books, etc etc.
It's crazy to me that we're still using the same approach to banking, given that the banking system regularly blows up and drags the rest of the economy into a recession.
The difficulty of this problem rests on the ambiguity of the phrase "exactly 6 months", which is going to depend totally on the precise business logic. But there's no reason to suppose that the requirements of the business logic will agree with the concepts implemented by the datetime library.
My personal rule for time processing: use the language-provided libraries for ONLY 2 operations: converting back and forth between a formatted time string with a time zone, and a Unix epoch timestamp. Perform all other time processing in your own code based on those 2 operations, and whenever you start with a new language or framework, just learn those 2.
I've wasted so many dreary hours trying to figure out crappy time processing APIs and libraries. Never again!
I'm also an American - I think this action makes us look weak and scared. Like our tech can't compete with Chinese tech. Strong nations act according to a set of principles, even if doing so is sometimes inconvenient or dangerous. Weak nations are constantly compromising their principles in order to survive.
I think to be complete this report needs to compare the levels of heavy metals found in the protein powders to levels found in regular foods like peas or brown rice, since that's where the metals are coming from. Is there some reason to believe that the protein isolation process would tend to concentrate heavy metals to a higher degree than is found in the source foods?
If you decide to go the low-tech/DIY route, I will shamelessly self-promote my platform WebWidgets.io. It basically allows you to upload SQLite file(s) and then talk to them in JavaScript from a web app. I built a finance tracker app for myself using WWIO (I do a monthly download of my spending data from my online bank and then upload the CSV to my app). I've been using it for years.
The key issue in this world is that everyone wants a slightly different set of features, so any one-size-fits-all app will be too complex. DIY-ing it means you get exactly the features you want and no gratuitous complexity.
I use Java extensively for my personal work and never use Gradle/Maven/Ant. I run Python scripts that marshal javac and java calls, it works great. I do cute tab completion tricks that examine the class layout and extend based on what it finds. I even have a "SmartCompJava" script that finds .java files that have newer modtimes than their corresponding .class files and just compiles them - a bit dangerous if you're not careful, but it's super-fast!