HNHacker News
TopNewBestAskShowJobs

unoti

3,471 karma · joined October 22, 2010

Applied AI engineering manager at Microsoft, our team specializes in cloud infrastructure automation: making the cloud efficiently build, grow, and heal itself.

Today I use AI to help businesses feel better, and previously I used AI and games to help people feel better. Formerly a game developer, virtual worlds, specializing in reuniting people with forgotten happy emotions and making AI critters that feel real enough to pay for(Mossms.com), and before that worked at Second Life. Before that I worked on ERP software for Fortune 1000's.

https://github.com/unoti/

submissionscomments
unoti··on Netflix buffering issues: Boxing fans complain about Jake Paul vs. Mike Tyson
> 1. Always be building: It does not matter if code was not changed...

> 2. Always be releasing...

A good argument for this is security. Whatever libraries/dependencies you have, unpin the versions, and have good unit tests. Security vulnerabilities that are getting fixed upstream must be released. You cannot fix and remove those vulnerabilities unless you are doing regular releases. This in turn also implies having good unit tests, so you can do these builds and releases with a lower probability of releasing something broken. It also implies strong monitoring and metrics, so you can be the first to know when something breaks.

unoti··on Please stop the coding challenges
Actually I put in that work myself, while working, and while raising kids. It took me a lot longer than a month or two. It took a ridiculous, embarrassing amount of time.

> And those I've met were absolutely insufferable and incompetent.

I understand this last part, and annoying and insufferable people absolutely do exist. But consider that you may be generalizing and discriminating against a pretty big pool of people using a small sample size, and give some of those people a chance-- as long as they're not being insufferable and are treating you the way you want to be treated. Those people exist in FAANGs too.

unoti··on YC is wrong about LLMs for chip design
>The problem with this extrapolation is that the software industry is pretty much unique in the amount of information about its inner workings that is publicly available for training on... millions of lines of code that we published on the internet...

> Nearly every other industry (with the possible exception of Law) produces publicly-visible output at a tiny fraction of the rate that we do.

You are correct! There's lots of information available publicly about certain things like code, and writing SQL queries. But other specialized domains don't have the same kind of information trained into the heart of the model.

But importantly, this doesn't mean the LLM can't provide significant value in these other more niche domains. They still can, and I provide this every day in my day job. But it's a lot of work. We (as AI engineers) have to deeply understand the special domain knowledge. The basic process is this:

1. Learn how the subject matter experts do the work.

2. Teach the LLM to do this, using examples, giving it procedures, walking it through the various steps and giving it the guidance and time and space to think. (Multiple prompts, recipes if you will, loops, external memory...)

3. Evaluation, iteration, improvement

4. Scale up to production

In many domains I work in, it can be very challenging to get past step 1. If I don't know how to do it effectively, I can't guide the LLM through the steps. Consider an example question like "what are the top 5 ways to improve my business" -- the subject matter experts often have difficulty teaching me how to do that. If they don't know how to do it, they can't teach it to me, and I can't teach it to the agent. Another example that will resonate with nerds here is being an effective Dungeons and Dragons DM. But if I actually learn how to do it, and boil it down into repeatable steps, and use GraphRAG, then it becomes another thing entirely. I know this is possible, and expect to see great things in that space, but I estimate it'll take another year or so of development to get it done.

But in many domains, I get access to subject matter experts that can tell me pretty specifically how to succeed in an area. These are the top 5 situations you will see, how you can identify which situation type it is, and what you should do when you see that you are in that kind of situation. In domains like this I can in fact make the agent do awesome work and provide value, even when the information is not in the publicly available training data for the LLM.

There's this thing about knowing a domain area well enough to do the job, but not having enough mastery to teach others how to do the job. You need domain experts that understand the job well enough to teach you how to do it, and you as the AI engineer need enough mastery over the agent to teach it how to do the job as well. Then the magic happens.

When we get AGI we can proceed past this limitation of needing to know how to do the job ourselves. Until we get AGI, then this is how we provide impact using agents.

This is why I say that even if LLM technology does not improve any more beyond where it was a year ago, we still have many years worth of untapped potential for AI. It just takes a lot of work, and most engineers today don't understand how to do that work-- principally because they're too busy saying today's technology can't do that work rather than trying to learn how to do it.

unoti··on Hackers use ZIP file concatenation to evade detection
> coders discovered that scanning the ZIP local entries to be a faster way to build up the ZIP archive entries, otherwise you're forced to seek all the way to the end of a ZIP archive and work backwards to locate the central directory

Would this have worked? Reserve a few bytes at the beginning of the archive at a fixed location offset from the start, and say "this is where we will write the offset to where the central directory will start." Then build the whole archive, writing the central directory at the end. Then seek back to that known offset at the start of the file and write the offset to the central directory. When creating the archive, we can write the central directory to a temp file, and then append that in to the end of the file we're building at the end, and fix up the offset.

Seems like this strategy would enable us to both have a number of files in the archive that are known at the beginning, and also allow us to do a high-speed seek to the central directory when reading the archive.

I imagine people thought about this idea and didn't do it for one reason or another. I can imagine why we didn't do that for Unix TAR-- most tape devices are a one-way write stream and don't have random access. But ZIP was designed for disks; I'm curious why this idea wouldn't have solved both problems.

unoti··on Please stop the coding challenges
> A conversation will tell you much more about their approaches to problems solving, team work, and mentoring than a coding challenge will.

Having hired many developers over multiple decades, let me assure you that you should definitely make sure they can actually code. Do definitely have those conversations, too, but trust me when I say that a technical conversation, a degree and several listed years of experience as a software developer on a resume are not enough to guarantee that they can actually code.

unoti··on Please stop the coding challenges
> Not that I'm complaining. I'm happy to pick up people that are good at computers but wouldn't be able to pass that hurdle, and probably wouldn't hire anyone that has.

You're looking for people who feel like they can't be bothered to learn some of the science and fundamentals of their craft? And you'd actively oppose hiring people who have the determination to go the extra mile? You'd actually discriminate against people who know how to use basic data structures and algorithms, and have that enriched landscape of knowledge to apply to problems that might come up?

Fortunately for you, there are lots of these people available to hire who felt like they are too good to put in the work.

I'm not saying that the data structures and algorithms knowledge is needed to do most software engineering jobs on a daily basis. But for lots of jobs, hiring people with a demonstrated willingness to dive in deep and learn things that aren't necessarily easy or fun actually can be a very good thing, because a lot of engineering problems require a similarly difficult dive into some aspect of specialized domain knowledge.

unoti··on Please stop the coding challenges
> Hiring processes should focus on problem-solving, collaboration, and growth in relevant areas.

Agreed. When I think about the best engineers I've worked with-- and the worst-- I have identified certain characteristics that I want to look for in a candidate, and certain characteristics that I must avoid.

In particular, there are certain people who just can't entertain any thought or design idea that they themselves did not have. This makes them very hard to collaborate with, except in the cases where what we need to do just happens to coincide with what they are thinking. Related, there are certain people who can only think about exactly one way to solve a problem, and there is not room in their brain to honestly consider any alternative approaches. These people tend to be really bad team players. A good engineer will understand that there's no one right way to do certain kinds of things, that everything comes with tradeoffs.

During interviews I present coding problems where there is more than one way to attack the problem. Whatever approach the candidate does, I propose an alternative approach and ask them to talk with me about the pros and cons of the alternative approach. I also ask for them to extend upon the alternative approach and make it better. For a significant fraction of candidates, they just can't entertain any other way of doing the thing, there just isn't room in their brain to consider alternative approaches.

Ability to really listen and give an honest consideration for other people's ideas is a vital part of the kind of team culture we need to succeed, so I look for that in interviews. Long ago I abandoned worrying about leetcode in interviews, and nobody in more senior management has made me stop yet, and this is at a big tech company.

unoti··on Chain-of-thought can hurt performance on tasks where thinking makes humans worse
> The argument isn't that there is something more than next token prediction happening.

> The argument is that next token prediction does not imply an upper bound on intelligence, because an improved next token prediction will pull increasingly more of the world that is described in the training data into itself.

Well said! There's a philosophical rift appearing in the tech community over this issue semi-neatly dividing people between naysayers, "disbelievers" and believers over this very issue.

unoti··on Chain-of-thought can hurt performance on tasks where thinking makes humans worse
Go back to the ReAct paper, reasoning and action. This is the basis of most of the modern stuff. Read the paper carefully, and reproduce it. I have done so, this is doable. The paper and the papers it refers to directly addresses many things you have said in these threads. For example, the stochastic nature of LLM’s is discussed at length with the CoT-SC paper (chain of thought self consistency). When you’re done with that take a look at the Reflexion paper.
unoti··on What Kind of Writer Is ChatGPT?
Regarding cost, doing something like this would be fractions of a penny. Obviously, the person doing the API calls is either paying for it, or they're paying for the electricity if they do it on their machine. But the cost is ultra negligible; certainly cheaper than it would be on Mechanical Turk or Fiverr. In fact so much cheaper that economically it wouldn't be feasible or worth the effort to try outsourcing it ordinarily. This is part of the game changer nature of AI.

Regarding blowing out context windows, yes, probably, but this is what loops and code are for. Think of implementing a system like a guided seminar that steps a person doing this work through it step by step, and giving them time and opportunity to iterate on and improve the product.

For example, with making up the D&D puzzles. Ask a college educated human to do this. You will find there's things you like and don't like about their results. Tell them more about what you're looking for, what you like, and what you don't like. Give them examples of what you like and don't like. Take notes on the things you discuss with them until you figure out a way to express how to coax out of a fresh person new to this topic how to give you what you want. When working with the person, work out a process where they have rough drafts, and you walk them through how to select the best items and give them pointers on how to improve on them. Write up a written process for how to do this. Maybe there are multiple phases in this process, and things go through multiple revisions to get to quality material. Now do the same thing with the LLM, and you have yourself a good system.

Same thing goes for writing stories, which elsewhere in these threads people say LLM's are terrible at. Sit a human down and tell them to give you a story, and I promise you will receive terrible results or outright copying. Instead, give your humans some guidelines. Like start with the idea that in the end our hero is going to be a particular way with particular strengths they need to conquer the central challenge. But in the beginning, they are the complete opposite of that thing. What is the central challenge, and what are the characteristics they need to conquer it? In what ways will the main character be the opposite of that at the beginning of the story? Then we put the character through hell in various ways over the course of the story to enact those changes in their character that they need to win in the end. For each of these sentences/phases above, make things that explain the ideas more fully, and make a process to iterate on these things, possibly with multiple different prompts and loops at every phase. This kind of approach more closely resembles what many real novelists do, iterating on ideas, often in the back of their mind or subconsciously, over hours or years of rumination with or without written outlines and notes. Maybe randomly select 2 things from movietropes.com and throw those in there saying incorporate these ideas. Experiment, iterate, see what works and what doesn't.

People need to give LLM's the domain knowledge and capability of rumination to succeed in so many of these domains, rather than just asking "write me a novel" and being disappointed. Or asking "write me a puzzle my D&D group will enjoy" without going through these extra steps that are implicit/intuitive for what experienced subject matter experts do.

Source: I write AI products for a living with many things in production delivering real business value at scale every day. It's not all hype, it just takes a while to implement.

unoti··on Do AI companies work?
> Anthropomorphizing algorithms only serves to impair objective analysis of when they are, or are not, applicable.

Actually, in this case, comparing how we as humans think to how LLM's work is in fact useful. It's hard for us to eyeball a word and say how many consonants are in it, we need to count. I wouldn't ask a human to eyeball a tax return and tell me what the totals are reliably without giving them the tools to add things up. LLM's are the same way.

It's true that anthropomorphizing in general can be a trap, but when working with LLM's it can be a useful guide in pointing the way towards workable solutions.

unoti··on Ask HN: Any good essays/books/advice about software sales?
Second vote for Spin Selling from me. Spin Selling is a must read for anyone doing long-term sales-- in particular, selling software that has a long sales cycle like a year from the time you get a prospect until you close the sale. But it has other key concepts for smaller software packages too that you'll find useful if you're doing something smaller.
unoti··on Xkcd 1425 (Tasks) turns ten years old today
A customer service agent isn't a bullshit job. They form a user interface between a complex system and a user that isn't an expert in the domain. The customer service agent understands the business domain, as well as how to apply that expertise to what the customer wants and needs. Consider the complexity of what a travel agent or airline agent does. The agent needs to understand the business domain of flight availability and pricing, as well as technical details related to the underlying systems, and have the ability to communicate bidirectionally comfortably with the customer, who knows little or none of the above. This role serves a useful purpose and doesn't really qualify as a bullshit job. But in principle, all of this could be done by a well-crafted system with OpenAI's api's (which others in these threads have said are "just chatbots").

Interfacing with people and understanding business domain knowledge is in fact something we can do with LLM's. There are countless business domains/job areas that fall into the shape I described above, enough to keep engineers busy for a real long time. There are other problem shapes that we can attack with these LLM's as well, such as deep analysis on areas where it can recommend process improvements (six sigma kinds of things). Process improvement, some might say, gets closer to the kinds of things Graeber might call bullshit jobs, though...

unoti··on Ask HN: Who is hiring? (August 2024)
Microsoft | Redmond, WA | Senior Software Engineer

Microsoft is hiring a senior UI engineer to help us leverage AI to streamline building and maintaining our datacenters at the center of our cloud strategy. The demand for datacenters has been high for years, and recent developments in AI has intensified this to unprecedented levels. If you were to tell ChatGPT to imagine we need to triple the world's datacenter capacity over the next few years and ask it what challenges we would face, you'll see a list of the things our organization is working on covering a range of issues including supply chain, land acquisition, networking, sustainable energy, repairs, and much more. Our team makes it easy for other teams to integrate AI into their own services. We work with a mixture of LLM's, small language models, and more conventional AI models to supercharge what our datacenter technicians and other people do every day.

Although this job offers flexibility with up to 100% work from home, you will need to be within a 2 hour commute of Redmond, WA, or one of these other central hubs in the US: Atlanta, GA; San Antonio, TX; Redmond, WA; Phoenix, AZ.

Job posting:

https://jobs.careers.microsoft.com/global/en/job/1750811/Sen...

unoti··on I prefer rST to Markdown
For the motivating use cases, I wonder if YAML and a simple script to transform to other formats as needed would be more effective, easier to read, easier to write, and easier to work with in general.
unoti··on Nvidia hits $3T market cap on back of AI boom
> Do you think that AI is having a big enough impact already to generate this valuation? Is it changing your life?

Yes, but the thing is: it takes time to write the software. We had proof of concepts for some serious game-changing things written and working over a year ago, but organizationally it takes a long time to write and deploy the things and get them into production. Even if LLM's didn't get any better than they are right now, there are still years' worth of ground-breaking improvements we could roll out but those will take time to write and deploy the software and reshape the business processes.

unoti··on Managing mutable data in Elixir with Rust
When they say all the data in Erlang/Elixir is immutable, it does not mean that there is no state. There's definitely data in the programs that change values, because like you point out, how in the heck can you write a useful system that doesn't have any state anywhere?

State or data that changes values is typically put in one of 3 different places:

1. On the stack. It's pretty typical to have a function like handle_message(app_state, request). The current state of the app is in the call parameters. At the end of this message, handle_call() would call itself again with the new updated state. Somewhere else in the system we keep track of the last value of that state, and if handle_call crashes, we just use the last state.

2. Another place to hold state is in external storage somewhere.

3. The third main place to hold state is via references to other objects, which do #1 or #2 above.

Regarding whether there is a difference between messaging local versus remote objects-- there's an operator for sending messages to another object. It works the same for local and remote. I think it's possible to inspect the actor address and see where it is, but the mechanism works the same.

unoti··on Managing mutable data in Elixir with Rust
> Immutable data is not a “foundation of scalability and robustness”.

It may not be the only way to get to scalability and robustness. But it certainly is the cornerstone of how Erlang gets there.

1. First, the way Erlang treats data ensures that every piece of data can be sent over the wire by default. This helps pave the way for another amazing characteristic of Erlang, and that is when you refer to and use an object, it's essentially transparent to your code whether that object is on this machine or another machine in the cluster. This would not be possible without the fact that all data structures are remotable, which is enabled by the immutable data. (See also side note below.)

2. The immutable data also leads to clean rollback semantics, making it easy to always have a self-consistent state of the system ready to use even after some kind of fault.

3. The immutable data also leads to very clean and easy ways to handle multithreading because you never have to worry about making object copies. You can be assured that it's ok for two threads to use the same memory object because there's no way either of them can change it.

Side note: Alan Kay, the inventory of OO, has said that people get the entire idea of what he was talking about all wrong. He said that object orientation isn't about objects, but its about communication. He was talking about the idea of an object being more like what we'd call a web endpoint today, where when you instantiate it you communicate with it by sending it messages. It's funny to me that a functional language like Erlang best embodies that OO idea today. Go code can, too.

"I'm sorry that I long ago coined the term 'objects' for this topic because it gets many people to focus on the lesser idea. The big idea is 'messaging'" - Alan Kay <https://en.wikipedia.org/wiki/Alan_Kay>

He goes on in the original underlying document to say "OOP to me means only messaging, local retention and protection and hiding of state-process, and extreme late-binding of all things." All of these ideas are front-and-center in Erlang (and by extension Elixir).

unoti··on Markov Chains are the Original Language Models
It's true that Markov chains are very limited in their capabilities. But one thing I love about them is that they are one of the simplest and most intuitive ways to write code that *learns* from input data.

If you're never written something that *learns*, try it out! Here's a very primitive one I wrote recently to explain the basic idea and explains it along the way.

https://github.com/unoti/markov-basics/blob/main/markov-basi...

This starts with generating US city names, and ends with generating text based on the Dungeons and Dragons DM guide.

unoti··on I don't want anything your AI generates
While arguing against the usefulness of ML models, the author of this piece is not thinking of two key concepts. 1) Supply and demand are linked. If the supply is cheap/plentiful enough, this opens up new demand for things that would have been prohibitively inaccessible before. To illustrate, imagine CPU chips were so cheap that they were essentially free, where the most expensive thing about them was shipping them. In that case, new demand might open up to begin using them to pave roadways or something. 2) There is knowledge and wisdom wrapped up in LLM's and recommendation systems that comes not from the computer program, but from the knowledge and wisdom of the crowds. This knowledge shouldn't be dismissed as useless out of hand just because it's in a computer.

The argument that computer generated art is worthless because "if you want custom art, pay an artist"-- this doesn't recognize basic facts about economics. In particular, supply and demand are related, they are almost the same thing. If the cost of free art is exceedingly low, this enables me to make custom art for things that I wouldn't even consider if the price or barrier to entry or time-to-generate were higher.

The author also points out:

> I don't want music recommendations from something that can't appreciate or understand music. Human recommendations will always be better.

While it may be true that humans understand music better than computers, the intelligence behind recommendation systems is actually not the computer programs themselves. It's the wisdom of the aggregation of many people, reflected in the numbers of the model. While talking with an expert curator about the next movie/series you should watch, or the next band you should discover would be nice, the truth is it's just not as practical or easy as firing up a computer recommendation system. While I could book an appointment with an expert music curator and go through a 2 hour interview process to find my perfect music tastes, a recommendation algorithm can still provide easy value for me.

unoti··on Ask HN: Who else is working on nothing?
If you're feeling like you're stagnating, here are some ideas to help get you jump started that I've done over the years.

There are a lot of things you can learn that are inherently *fun*. Maybe try learning something that sounds fun and interesting to you! Some examples for myself include things like Ham Radio, how and why radio signals bounce off the atmosphere. Or cooking, figure out how to make the best salsa you've ever had by getting a molcajete and fire roasting some tomatillos and peppers. Or go learn Unreal Engine, and the endless wonders and rabbit holes therein. Or make your own toy programming language, or teach people about programming, or make it a goal to make the best apple pie, or chili, (or insert your favorite food here) that you've ever had. Learn how to make the best margaritas the world has ever seen. Or take up making homemade ice cream. Get into 3d modelling or animation or texturing. Take up dog training, and do agility or dog dancing. Take up camping, hiking, backpacking, cross country skiing, mountain biking. Learn about wilderness survival, or backcountry emergency medicine. Getting certifications can be a fun way to force yourself to learn things you wouldn't otherwise, and my ham radio license and emergency medicine first responder certifications were really fun to get. Learning to draw can be super fun, and easier than you might think (get the book "Drawing with the Right Side of the Brain"). Fitness, take up some form of exercise that sounds fun to you; in my case I did powerlifting. Take the Fastai course and learn to make ML models from scratch. Learn to fly the A10 Warthog in DCS-- they say that if you can do that, you can do anything in simulation gaming. I'd go further than that and say that learning all the content in the FastAI course was easier than attaining a level of mastery with the A10.

People skills: One of the most impactful things I've ever done is take the Masterclass in Negotiation. That and reading the book Nonviolent Communication changed how I think about and deal with people forever. The book on body language, What Every Body is Thinking is very fun and will help make you never be bored in a meeting again. The Like Switch is also pretty great. Together I consider these a pretty fabulous two semester course in dealing with people.

Getting started is always the hardest part. Once you get some momentum on something it gets easier. A lot of these things can be combined, like you can listen to tutorials about how to fly the A10 while you're out walking.

unoti··on I pwned half of America's fast food chains simultaneously
I worked with Firebase for a while, lured in because of how easy it was to do certain things. It makes certain kinds of operations essentially zero effort, such as getting realtime updates on the frontend when something changes. But it also creates a huge amount of effort that is trivial with other frameworks, such as creating a huge effort for security. I found that what I gained in convenience, I lost by needing to do so much work continuously battling with security rules. I left it behind and never looked back, and it made me much more cheerful about the work that I needed to do to establish and maintain more conventional backend data systems.
unoti··on Write your own retro compiler
For me the interest in retro computing is that the computer is so much simpler that it is indeed possible to understand. The set of assembly instructions is small enough that a person can understand every opcode fairly easily, the amount of memory is limited, and the list of things that that your installed ROM can do is relatively accessible. Contrast that to modern machines and operating systems, and no mere mortal can understand everything that is happening in the machine.

Another thing to love about retro computers is how transferrable the knowledge is to modern machines. Once you know the essence of assembly language on a retro computer, you have a good basis for learning RISC-V or something else modern.

Depending on where your son is at, he might enjoy this video where I explain the essence of assembly language-- it's the video I wished someone had shown to me when I was a kid and thought I wasn't smart enough to learn it.

https://www.youtube.com/watch?v=ep7gcyrbutA

unoti··on Accidental database programming
One other piece of prior art is Lotus Notes, in which the database was stored locally and synced in the background. https://en.wikipedia.org/wiki/HCL_Notes
unoti··on Is anyone using self hosted LLM day to day and training it like a new employee
> What I would love to get working at some point is allowing an LLM access to my schedule, notes and goals and then have it help prompt me at appropriate times. "Hey, TJ I noticed you haven't worked out this week, it's sunny today this might be a good time". That sort of thing.

If you work in a Microsoft world, this is what GraphAPI is all about: enabling access to all the things using your personal authentication token. This includes emails, calendar, OneNote, One Drive, and essentially everything. I've been working on making an easy to use OneNote provider with GraphAPI underneath it that I can use to work with the LLM.

https://learn.microsoft.com/en-us/graph/overview

https://learn.microsoft.com/en-us/graph/use-the-api

unoti··on Robot Dad
Eleven Labs is blocked by my organization at work. I don't know why. But I did find that Azure has a way to train custom text to speech models here https://learn.microsoft.com/en-us/azure/ai-services/speech-s...

I'll also check out some of the other services described in sibling comments.

unoti··on OpenAI staff threaten to quit unless board resigns
Awesome point. Microsoft's market cap today went up to 2.8 trillion, up 44.68 billion today.
unoti··on The Tech Gold Rush Is Over. Where's the Next One?
> Not sure where the next revolution may come from (no, 'AI' is not it).

Not sure why you say this; are you sure AI can or should be so easily dismissed out of hand without further thoughts?

---

The assertion that "AI is not it" when considering the next revolution is a point of debate. While it's true that AI hasn't yet led to a socio-economic transformation on the scale of the internet or mobile technology, it is still in the early stages of its potential impact. AI and machine learning are already influencing many industries, from healthcare with predictive diagnostics to transportation with autonomous vehicles, and their capabilities are rapidly advancing.

The part below "---" above was written by ChatGPT, to help underscore my point.

unoti··on The AI research job market
rg111 has a good strategy here. For step 1, making a cats vs dogs classifier, the first lesson of the FastAI course will give you a path for doing this right away, and I wholeheartedly and enthusiastically recommend this course.

A sibling commenter mentions that certifications will do nothing for you. They're not exactly wrong, because what ultimately matters is that you can demonstrate your skills. Certifications and to a large extent even degrees mean very little; what matters is that you convince them you know how to do stuff. The best way to convince people you know how to do stuff is to be able to show a list of cool things you actually did. These courses and their certifications may not mean much on their own, but in the course of completing the courses you will develop skills and capabilities you can demonstrate and talk about in your resume and cover letter.

unoti··on The AI research job market
> a lot of places are stuck on the idea if you didn't do it in the past you cant do it now

The trick for breaking into something like this is to produce a portfolio of one or more projects you did where you demonstrate experience with it. This means actually doing it yourself, have a repo with notebooks and text explaining how everything works.

AI is definitely not the first field that is like this, where it at first appears only the people already doing it are qualified to do it. I have had to do this quite a few times over the last 30 years to stay relevant. It takes a lot of work to do this, but it's easier than ever to do today. Today the tools you need to break into almost any technical field can be freely downloaded. A couple of decades ago if you wanted to create a portfolio for something the tools were not freely available. For example, vxWorks for embedded systems programming, or Oracle for demonstrating you can administer large databases, or 3D Studio Max or Maya for 3d modelling: all of these tools were expensive enough to be inaccessible to an individual.

But today, you can go do independent work, take courses and get certifications, and create your own body of work that demonstrates you have an understanding of the field.

If you want to start making your own body of work in the field of AI, I suggest starting with these resources:

1. FastAI

2. Deeplearning.ai. Get a certification and put it on your resume.

3. Karpathy Zero to Hero

4. Re-create the technique in the ReAct paper (Reasoning and Acting).

If you want to demonstrate capability in AI in general, proceed in sequence 1-4. If you want to demonstrate capability with LLM's in particular, proceed in sequence from from 4-1.

← PreviousPage 2 of 27Next →