To be fair, even reading 'good' books won't make you smart. I think the key is to be critical, which should be thought at a young age. Ikram Antaki dedicated most of her last years in teaching this in Mexico.
Anecdote: When I started studying economics I really agreed with a lot of what I read from economists like David Ricardo, Marx, Smith, etc. Then, I studied what other economist had to say and I could see how they disagreed with the former. This made me realize that I agreed with those people because their arguments 'made sense' to me, but that doesn't mean that what they said is completely true. This is something that has stayed with me, I always wonder how can something be wrong.
Long ago we abstracted programming into a logical language which allow us to think at a higher level. IMO LLMs are another abstraction but a bad one as it is stochastic and we can't guarantee output quality (e.g. security, performance, etc). The dream has always been to tell the computer what to do in a simple language, and the challenge has always been finding out that we didn't even know what we wanted the computer to do. LLMs might help in the first one but not in the latter. At the end, human intelligence cannot be outsourced.
The idea of offshoring computing is good. However, the cloud developed as a centralized computing platform instead of a distributed one. This has created power dynamics that harm customers. The same happened with social media, and has happened to other industries. I think it would be better for customers if there were many small cloud providers and they could easily switch between them. But even migrating from one cloud provider to another is a huge endeavor these days.
I think AI labs are realizing that they no longer have any competitive advantage other than being the incumbents. Plus hardware improvements might render their models irrelevant for most tasks.
I don't really think it is turning into a guesswork. A lot of people wrote bad code before by pasting things from the internet they didn't understand. I think some people are using LLMs the same way, but it does not mean that programming has changed. But I do think that code quality is being neglected nowadays.
What does an AI degree provide? Is it really different than majoring in CS? AI has been pretty standard in CS undergrad programs AFAIK. At least in high school it seems that AI curricula is just learning to use LLMs and understand a few concepts. That does bring value at a society level but I am not sure if this makes sense as an undergrad degree (if that is what they are only teaching).
I do agree with his identification of the problem: sometimes agents fail because of the tools around it and not because of the model's reasoning. However, for the failing tests I think he is not making the distinction between a failed test due to a harness failure or due to a reasoning failure. It would be nice if someone analyzed that from the data set.
I don't think so, have you read 'The Bonobo and the atheist'? Humans are not the only ones using tools and in reality there isn't much difference between humans and animals. The conclusion I get from the book is that the only difference is religion. Although, I have a feeling that humans do have a more developed intellect (problem solving) but this was not explored in the book.
Yeah, there is also the work of primatologists which challenges some of our beliefs of what we think is human sciences (like politics). See Frans De Waal.
Yet, I believe there hasn't been much progress as compared with STEM. But it is just a belief at the end of the day. There might be some study about this out there.
They usually require more data It is not a great idea to diagnose anything with so few information. But in general I am optimistic of the use of LLMs on health.
I agree. Most of the time people think STEM is harder but it is not. Yes, it is harder to understand some concepts, but in social sciences we don't even know what the correct concepts are. There hasn't been so much progress in social sciences in the last centuries as there was for STEM.
That is very interesting. It is funny to see how influential the federal government has been on society, infrastructure and other areas of life. Specially considering that some people opposed to it during the confederation period because they saw it as another centralized authority (anti-federalist papers).
Transportation influences urban development. That is why most houses have a garage. There is no such thing as private transport (streets are public). Transportation has been heavily centralized since the New Deal. The bicycle was okay for most people living in cities in the 30s, now it is not because the government has favored the car infrastructure over the last decades. I think we need to start with not letting government develop their big infrastructure projects which are not resilient. Advocating for the use of bicycles might make sense in some places yet bicycle infrastructure is required.
I find it very useful for code comprehension. For writing code it still struggles (at least codex) and sometimes I feel I could have written the code myself faster rather than correct it every time it does something wrong.
Jeremy Howard argues that we should use LLMs to help us learn, once you let it reason for you then things go bad and you start getting cognitive debt. I agree with this.
I think LLMs are helpful for understanding code. I used to spend like an hour trying to find where something very specific was made, and now I can just ask an LLM and it finds it right a way and is able to explain how the code works. This is probably the thing that has saved me most of the time.
I agree with the point that learning requires work. In general, everything worth doing requires work. This is one of the things I often have to remind myself, otherwise I spend the whole day 'learning' and I just read a bunch of stuff online that I then forget, instead of trying something out which I actually will learn and understand.
The robotics thing to replace caregivers misses the point that elder people also want connection. Yeah, it might free caregivers but still we will have a loneliness epidemic. I think this is more related to the desire for progress which is the backbone of modern life (you see it politics, school, your family, etcetera). This, I believe, has been slowly replacing the social glue of societies like religion, public space, play, chatting, etcetera.
As someone else mentioned here: not all technical debt is created equal. I agree, sometimes the problem are changing requirements, etc. But it is also true that there is technical debt caused by developers who don't take the time to properly design features and will simply implement the first thing that came to their minds. I agree with the author, this kind of technical debt is caused by a mediocre attitude which often propagates to all the team if there is no one that calls it out.
The more interesting discussion to me is: how do you solve this problem once it exists in a team? I guess there are many approaches, but I tend to think that 'lead by the example' is the best you can do as an engineer, but a top-down approach might work better which is what happened at Microsoft when Satya Nadella became CEO.
> "I expect us to go back to extending our agents with the most accessible programming language: natural language."
I don't agree with this. Natural language is so ambiguous. At least for software development the hard work is still coming up with clearly defined solutions.
There is a reason for why math has its own domain specific language.
I was thinking the same thing. Maybe is that at the end the author seems to imply that agentic AI will work simply because models have become better regardless of the way we make them agentic (i.e. MCPs, skills, etc).
I think this is purely first mover advantage. We get stuck with bad products simply because those were the first products on the market. It is difficult to change them once everyone uses them. The same applies to the adoption of IT on the banking industry. Now we are stuck with COBOL and systems that are hard to migrate without damaging the economy.