Use this link to check out interview kickstart, and not waste your shot on interviews www.interviewkickstart.com/su... / aaronjack #coding #hiring #programming
AI is not just when it is LLM and Deep Learning. You could of course come up with a solution which reads, parses and evaluates the code. This would communicate with the other part.
I have reply to the NVIDIA CEO: IMHO is easier to create AI that replaces upper management than to make an AI that replaces a senior SW dev specialist. Maybe we should start working on that ?
@@Lisekplhehe Just like the devs that are developing AI to replace fellow devs. You'll always find short-sighted scum that is willing to take on a task of eliminating their competition for a short-term gain; maybe then even think they'll be granted immunity for the good work.
The high pay rates in the US are another factor. I retired (as a SQL developer) at the start of the lockdown (I was already well over 70, so I had all my bonuses) but the writing was on the wall. More and more of the coding where I worked was leaving US and heading to India and some low cost eastern European countries. If your job can be done from home, your job can be done from anywhere in the world.
This is how the bean counter finance CEOs think. The only problem is they discount experience, time zones, soft skills like communication skills, etc...
Yes, thats how it is and in best case we get the honor to fix the broken code later when the customers are tired of paying a «low price» and dont get anything back.
Are any American software engineers agreeing to lower salaries and starting move to remote locations with very low standards of living to deal with such apparent economic realities?
Thank goodness someone FINALLY talks truth about code. Literally every dump I get from ai I have to proof. May as well do it all myself from scratch. Garbage in garbage out.
95% of the time, AI creates code that just flat out sucks. You spend more time trying to get the AI to massage the code to get it right than you would have if you had just written the code yourself to begin with. I don't think LLM will ever fully replace software engineers and any attempt to do so will more than likely fail.
Actually it wont. The reason is because if it's a broken metric, and poor candidates start passing, then the tool will be abandoned / lead to company failure. "Bad money drives out good"
What degree did you get? You can still be a plumber. Unless you are running a plumbing repair business, work can be pretty sketchy, hoping for new construction which is quite variable.
I remember I called a plumber at 10pm to replace a hot water heater tank that broke. It couldn’t wait since we were going to have guests and my wife did not want to take a cold shower before work. I saw the plumber arrive and man handle a 200lb water heater down into our basement and man handle the old 200 pound water heater back up the stairs. I remember thinking this was tough work. I don’t think I could physically do what he just did. Could I do this when I am 50 or 60? I don’t think so. While plumbing pays pretty good it’s a tough job and you need to be physically fit to do it, which you can’t count on. I hurt my knee playing sports and if I was a plumber that means I wouldn’t work for several months. But since I had a desk job I was still able to work.
There was another recent video (Brutal Truth Behind Tech Layoffs) that seems related to this first "cheap money" discussion. Basically tech firms had to justify all the money they received, which led to spending it by over-hiring and ultra-specialization such as "button engineers". And after reality set in (led by certain people coming into companies and slashing "unnecessary staff") this led to a wave of layoffs and a now-flooded market. It also pans AI as not-a-reason and not-really-a-threat, so there is commonality there. The tax code discussion is VERY interesting and something worth looking into.
And like in 2000 and 2008 the competitive market space is squeezing out the "self taught" mostly front end folks. I really hope people learn lessons this time around: You can get a job for a few years and make some nice mkney when market is hiring mediocre relatively uneducated devs en masse, but it's not a sustainable career for most people.
I’ve seen this happen at smaller manufacturing companies. They buy a lot of stuff when it becomes clear it’s gonna be a good tax year. The law was written for manufacturers like that in mind to incentivize them buying more machines and making even more stuff and hiring more people to do it. Doesn’t translate well to software. I’m sure all the billionaires knew this…Elon was the first one to just say f it and downsize. Most of those mid tier manufacturing companies sold out or offshored, but back in the day, taxes drove unnatural behaviors. A lot of those owners are looking to retire, not grow. The tax code keeps them in the game. I only knew one owner that just sucked up those losses (and that is definitely how they all view taxes, just like most people) for a few years to get across the finish line. A lot of them would be waiting. For a buyer that wasn’t just gonna liquidate, either. And those people are nonexistent now. The margins are non existent when competing with offshore slave labor. Almost no one wants that for themselves in the long haul, especially not someone who can raise the money to buy a place like that.
the system is designed to put working people down. the industry had it's honey moon period, now comes the time to enjoy near-minimum wage with the rest of the working class.
I used chat GPT for a while until my company forbade it. It was useful to explain Javascript programs to me, since I am mainly a C/C++ programmer. As far as using it to write programs from scratch, this is a management wet dream. It would be like firing all of your good programmers and hiring idiots, which is not far from what they are trying to do. Good luck.
That's weird. Now companies are banning chatgpt - they don't know what they are doing. They can't ban liquor, tattoes, and cigarattes. And they want to ban chatgpt, only speaks about their own work ethics.
this here. it's a good tool to help me understand new things and breakdown big codes to easily understand. the ai who can replace a full on developer is not here yet
Besides I'm sure that the guy that will craft the proper prompt will NOT be a manager. Now that means one guy can replace an entire team, but that also means projects can be made much faster, so likely more projects will be made at once
Try being more Indian. LENGTHS PEOPLE WILL GO TO IGNORE THE H1B VISA PROBLEM.... Each president brings in 300,000 H1B visa per year over 20 years is 6 million in an industry with 6 million workers. Who do you work for google?
@@powerHungryMOSFET it most likely will imo. I hope people adapt to use Ai for software development. Bc it's going to Inc our productivity and hopefully our salaries. Im certain it'll atleast stop jobs going overseas to India every company I've worked for developers overseas in India has been a disaster with there language barrier and there experience level was severely lacking to the point there counterproductive. Every h1b visa person I've been has worked harder than the average American here though so... You gotta compete with that
I can definitely confirm this, did my taxes and found out I cant write off most of my dev costs from last year. It's definitely more challenging to do a startup now. The cost of off shore development has gone up quite a bit which makes this risk even greater, especially if your not bringing in money yet.
Offshore dev is another red herring. The wave that started in 00s wasn't so much driven by cost saving. That was a secondary bonus. Outsourcing began and continues to be driven by dev ceilings being hit in many countries. There's only ever a certain amount of qualified devs in any given population at any given time. Once you reach that ceiling, you have to look elsewhere. That's been the salient driver for dev outsourcing in US for decades.
I am a bit confused, but maybe you can enlighten me. Can't you just by default discount the salaries of any of your employees on the revenues to reduce the tax burden? Like HR employees, marketing, etc?
Off-shore development works only when people know your kind of work. Soft dev work isn't like that and rarely you will get people from that pool. Once they earn enough, they want to become managers or try something else, not as taxing as dev work that takes over every part of your life, knowingly or unknowingly. Several psychological breakdowns will eventually led me to believe that human society isn't meant for developers.
@@iorekbyBrah... sorry to say it... but your giving us (American Tech Workers) wayyy too much credit! Trust me they could ship every single Dev Job overseas... and it would barely make a dent in Day 2 Day operations! Companies simply can't do that due to the Negative perception it would have on the Stockprice, hinder the ties with local Business communities, and most importantly... have them on the Wrong side of local & Federal Governments. We're Good...but so are a LOT of People! We ain't that much Better than everybody else!!
@@UFO_808 I take it you haven't heard about the StackOverflow drama, then ? I'll let you explore that one, it's edifying. As for myself, ever since I suspected my work could be used to train AI's, I've stopped posting code online. As a hobby, I also lie to any and every coding-related question I'm asked by strangers on the internet. I'm making sure the AI well is truly and completely poisoned so that AI code generation ends-up worse than useless. I'm not the only one doing this. Don't be surprised when the next ChatGPT writes you code that doesn't even compile. As for us becoming more valuable in the end, I'd say it'll be a one-time spike: right now I'm sure a lot of start-ups are busy putting AI-generated code into production that, it'll turn out, won't be maintainable but will be vulnerable. All that garbage will need to be audited and replaced by humans. We've had code generators and rapid-development tools for dozens of years, they haven't replaced anyone. AI is just worse crap but with a fresh coat of paint. Because at least code generators can be engineered to produce auditable, reliable, maintainable code.
that's really surprising ,now I understand why the job market is down in the last few months.I decided to move off from coding and do something else,not because of AI,it's just because chasing bugs in a 4 walls box is not fun anymore ,I'll do an outdoor work
Thank you for sharing this. I have dreams of launching a startup in the near future and this news is just heart breaking. It feels like they are actively trying to destroy the creative spirit by using taxes as the new shock collar for behavior adjustment. This has nothing to do with needing more tax revenue. It's a tactic to discourage and bankrupt creatives and get rid of future competition.
The kind of start ups that are impacted are the ones whose strategy is to rack in loads of venture capital and go for all out growth without profit. YES, the ship has sailed for now. HOWEVER, if like me your making a low risk, sensible, profitable company for yourself, you can 100% still do it, even now
@@ctb1977 Thanks for walking me through that. I have a better understanding now, and after further digging realized the rule isn't going to impact the project I'm working on.
@@ctb1977 Actually it's the opposite. Many times VC funding is not counted as income, so VC backed companies are technically operating at a loss. Break-even bootstrapped companies are the ones who get screwed.
If you want a job you should probably study full stack, deployment, and cyber security, it's called DEV SEC OPS. Like DEV OPS, but with the addition of cyber security in mind. I doubt I would get a job knowing css, html , and javascript.
@4115steve (Me--Captain Obvious...I know) for sure cyber security here for your children's children. Sadly probably for their children. Thanks for the tip.
The security industry people I'm connected to on LinkedIn are suggesting that there's effectively a hype bubble related to cybersecurity careers, and that the reality is that the field is not growing nearly as much as the hype suggests. In other words, it could end up as a dead end, a field that gets oversaturated with talent relative to the long-term positions it will actually offer. Not that I think there's anything _wrong_ with getting security training (it's a good idea on GP for anyone involved in tech TBH). But I don't think it will solve anything. I *have* experience in DevOps (*lots* of *professional* experience), and have worked in _literal_ tech security companies, as well as finance-related ones (which take security seriously, and expect at least some level of knowledge about it from their tech personnel). I've also worked as SWE and SDET. I still am not getting anywhere in the current market. Even with a clear track record of having learned new skills _on-the-job_ as needed (or just helpful), and brough them to bear for former employers. I'm not saying _don't_ learn DevOps, DevSecOps, or cybersecurity. I'm saying don't count on having learned them to change the narrow-minded, myopic approach to hiring that companies are currently engaged in. And we can blame ZIRP and the new tax code for their behavior, or we can look at this more reasonably and say that ZIRP was never reasonable in the first place, and encouraged and entrenched very bad behaviors within the tech industry, and that companies _not_ paying taxes (including startups, which have been the vast majority of my past employers) was never reasonable in the first place, and has left us in a world where multi-billion dollar companies essentially pay zero in taxes, while citizens, actual human beings, get no such break at all (you know, unless they're ultra-wealthy, and don't actually _work_ for their money).
Can you imagine having millions of dollars from selling GPUs but still prevent parents from teaching their kids a kill to help them put food on the table!!
If you actually read the full quote from Nvidia CEO he said the reason is because they would rather train a smart 20yr old newbie than a kid whose learned to code bad JavaScript from 57yr old Mrs Benson at secondary school
@@ctb1977smart 20 year olds probably coded when they were young and no child in the future has to learn from their teacher (solely) when we have the internet
@@ctb1977 I get this sentiment but I highly doubt Nvidia is hiring people with no formal training in STEM. You could argue the same for math or any subject. Anyway I don't think learning programming prior to high school is especially useful for most people since they're still working on being literate, thoughtful people that can easily figure out problems while reading documentation and self teach. I know a ton of people who went to colleges that aren't prestigious and have good dev jobs now making $100k/y where you learn it doesn't really matter, everyone is self taught eventually.
that final 5% is what has bankrupted some self-driving startups (we were told self-driving semis would be mainstream by now) and companies whom have poured 10s of billions into it (TSLA) have still not cracked.
Yea, imo theres still a lot of work to be done. AI in SEG companies is also a pressing topic as some people are falsely claiming that large language models can now distinguish the intent of the user
I use chat gpt a lot to help generate vocabulary lists for language learning (human languages, not related to coding) and I'm amazed at how much it struggles with basic tasks like not repeating the same word in a list, or making lists greater than 100 words, or making a list of purely adjectives (not mixing im verbs, nouns etc) and I kind of have a chuckle when I think that this thing in it's current form will take my job. It's a useful tool but it's so bad in certain situations even with reallt simple things
I would avoid using AI for foreign language learning. Building vocabulary lists via immersion is a much better route. You have to try to find immersion material that is slightly harder than your current skill level. Steve Kauffman, AJATT, and Refold here on youtube detail the process quite well.
I don't understand people cannot see the benefits. Before we were using just dictionaries and encyclopedia then encarta then Internet and now all that and chat gpt. They are just tools. I made incredible things to learn using chat gpt. Specially if you have social anxiety
@@colincotterell3365 Ive watched many of his videos and I respectfully disagree. Learning lots of vocabulary is also very important. I speak Russian and spanish fluently and lived in mexico for a year and have been in russia since November, so I'm familiar with immersion. But if you don't know any words or know very little then it's hard to understand people, I don't think that's very controversial or wierd
@@manoo2056 yeah I'm not saying theres not benefits. Just that it's useful but also has some issues in it's current form. Some of the mistakes it makes are so simple that I don't understand how such a complicated program struggles with these things. Like it really struggles with removing duplicate values, so much that I just do it manually after wrestling with it sometimes
@@Matt-jc2ml Yeah that's fair, AJATT/Refold have you memorize the top 1000 words in the language before you go out and do immersion. I don't agree with Steve on immersion at day 1.
I've seen this cycle so many times in my career. This is not live large in the cubicle time, or even join small startup time. You should be with your mates having ideas, doing things just because they interest you. If you are worried about interviews you are doing it wrong. And yeah, except for a certain romanticized version of the garage, it is not a really good time for people who are not garage startup kinds of people. And that does suck.
Amortizing R&D cost should be an option, not a requirement. I can see an established company amortizing some stuff, but for a startup, that's just impossible to deal with, unless you are doing it on your own, which has its own risks.
To put it simply: Images: A single error rate of say 1 in 1000 would probably be acceptable Software: A single error rate in 1 in a million lines of code, whilst it sounds ok would be unacceptable as that 1 error would be extremely hard to find and fix if you dont understand the whole code base. Also with embodied AI, if a robot was saying doing the dishes... a 1 in a million error would most likely be more correctable than the code example as the impact would be negligible.
The tax changes and tech de-bubbling are strategical, the US is trying to reindustrialize so these changes are probably there to make the intellectual part of the workforce flow into other fields. I think the tech bubble will be gutted more in the coming years. As for AI helping code, its limitations are very evident as soon as you try to make it do something non standard, make it use rarely used language features, or perhaps explore the possibility of using something that doesn't really exist. What I frequently get is instead of denying the existence of the feature it starts generating complete rubbish. This is when you realize the lack of intelligence in the thing and its true nature as a convincing spam generator. For programming tasks there needs to be an artificial general intelligence. If this is ever achieved it will also need to beat human in cost efficiency. The latter may be very uneasy given how optimized the biological systems are by billions of years of evolution. Try making a bird sized aerial vehicle do an intercontinental flight on a single energy charge. This is what a typical migrating bird can do. There may be a similar gap of artificial vs biological general intelligence if it is achieved.
The most difficult part is to figure out what the program. Coding is fairly easy actually. The value is not in writing the code, but in figuring out what it has to do.
What AI will create is huge amounts of code review, which should have been going on already but is lacking, and that won’t take long to get going. Also, AI may allow large companies to collaborate much more than they do now while still respecting each other’s patents allowing companies to pay fair prices for the use of other companies patent protected inventions.
AI currently say whatever you want hear, because you are supposed to correct their answer. This kinda put the horse before the cart, reviewer need to be able to point out submittant mistake. Basically change his view on how he operate. Chat bots would just change its answer everytime you protest them. Inconsistency at best. At worse? It would be like an echo chamber.
If the company is not allowed to write off wages but instead has to capitalize them, then they have to raise capital in order to pay the wages. They can't pay for them with income from the business. In your example where the company earned $100,000 and they had $100,000 of matching wages, they paid $40,000 in taxes. That only leaves $60,000 left to pay the wages. They would have to bring in outside debt and equity to make up the other $40,000. Essentially, they would have to use debt and equity just to pay taxes. That's insane.
Exactly. It totally screws over small undercapitalized outfits that can pay wages out of income in favor of VC-backed firms that make no money but can pay wages out of banked cash.
So you hire a developer as a janitor that has to sweep 1 square foot of floor a pay period and they make a huge hourly wage but only "work" 1 minute. Salary works out at the end. Oh and you also need a company policy that they must use a company broom to sweep and the only way to earn access to the broom is by developing software at no cost for x amount of hours a pay period. Im sure you can deduct maintenance like cleaning staff from taxes. Easy
18 months later and just *now* everyone is starting to say the same damn thing I was showing on twitch. Thanks! Glad more people are getting the message.
As a software developer in Europe: in the UK and Spain specifically, whilst opportunities are down from the 2021/22 peak, they haven't fallen off a cliff and as a senior dev I still get recruiters contacting me. But the salaries never were comparable with the US
AI can do a "give me a function that..." but it can't decide on designs, it can't give you a reasonable choice of what to use and how, this still required a skilled developer (team).
I recently decided to use Gemini for a couple months. One of the biggest limitations is I could only cut-and-paste 500 lines of text whereas with GPT-4 I was cut-and-pasting 2-3 files that were each 500-1000 lines. However the 1.5M token gemini is supposed to be rolling out to more and more people. Why are other people having trouble cut-and-pasting full on files? (9:14) The other limitations mentioned are definitely there. AI can't reason very well. It usually does better if I keep dozens of prompts in the chat history in the same conversation so it can predict my style a little better. It's great at writing test suites or toString styled procedures, add comments, or translating python syntax to rust or whatever. It's bad at thinking.
Man I am having the hardest time building a document processing and text embedding app for Gemini right now. Chroma DB and Lang chain documentation doesn't seem to agree with vs code
Spot on. The Feds are reversing interest rates and doing quantitative tightening to battle inflation. This sadly is also reversing the wild investments into Tech stocks and companies during covid. There will be point of stabilization in the tech industry and software engineers are still required at a company but it will not be the same as the peaks during covid. This means that you will need to develop your skillsets even further to stay in the job market OR pivot to a different industry that will net you similar or better gains
I agree about venture capital being an important source of money for IT. Another important source is companies requiring digital services. But many companies, at least in Europe, are strugling and can't afford digital services. They have more expenses for energy and employee salaries and other stuff because of inflation. Digital services is a bit of a luxury, so they'll just spend less on that. So, actually, two sources of income are reduced. Then there's also the weird trend of looking for candidates that have experience with only the specific tools that you are using, which is, it actually takes experienced developers about a week to learn a new tool...
especially machine learning! I'm a frontend dev who's starting to get into machine learning cause I believe that'll be just one of the most lucractive fields in the long run.
Exactly. Writing code is about less than 10% of (programmer/developer/software engineer/whatever)'s job. The rest 90% and something is reading someone else's code, or reading one's own code wandering: "What the hell was I thinking about 4 months (or more) ago. Now... AI doing the said 10% of writing... I guess that should keep our jobs for a while. On a personal note: I've punched my 1st program in high school, in the then FORTRAN, on Hollerith cards. Still happily programming at the age of 65, not worried about my job prospects in the slightest. Not FORTRAN programming, of course, thank goodness. And, when the end comes, when AI will actually be able to understand legacy code, and explain it to the mere humans - I'll probably be dead. I'm eligible for pension as it is.
this was a surprisingly well thought out topic. didn't know about 174, and of course it makes a lot of sense. I definitely use AI now as a tool to enhance my productivity and bridge knowledge gaps through asking lots of questions and getting a lot of relevant information that I can use as launch points into learning more. It's great at finding sample code that I can use to adapt or have a template explained but for complex scenarios it's definitely not useful at all.
Years ago I saw the specs for a gov project that contained a statement about software that writes software shall not be used to develop code for the project. I wonder if that’s still in place.
I also feel people who are self though and only doing front end / start up jobs it's going to be hard on them. This is why emphasize the need for a degree because you can get that advantage of a star up while also becoming way way more knowledgeable in computer base such as bit numbers, terminal use, other high-level language, and the in's and out's of a computer. This is something that boot camp website don't show you, which in my case is good because it removes competition for me. I dough anyone is going to read my comment and revolutionize it, but at least one person might see it and can make a good decision when it comes to what routine to choose in career.
I led engineering organizations. Competition in high tech means being at the cutting edge of applying new techniques and technology. LLM's, by their nature, lag behind new approaches because they are based on being trained on what has been discovered. Humans will always be the entities that find these new approaches. What it means is that jobs will favor those that can develop these new approaches.
Cal Berkeley Economics grad. The Sec 174 content is fire. My sons work in the California tech sector. They have seen the carnage from Sec 174 and ZIRP. Your video explains everything. Of course the 2010 housing crisis legislation generated ZIRP through two presidents, raising prime rates from near 0 to 6.5% was insane! This video is sending a not so subtle message to tech workers and companies about the future. And today’s assault on Apple may create an adversarial relationship between tech and politics.
Even as AI advances, there's still one issue I see with coding... wrapping up a project right now, it's nothing complex but lots fo smaller parts, but in order for this to be successful, I need all those small details in there and done properly. Even with a more advanced AI that can actually write code, more than likely it would take me longer writing never ending prompts to get all these small details hammered down versus just writing the code myself. I could just shrug my shoulders and run with whatever the AI gives me, lessening the chances of the project's success due to a lower quality, less custom tailored end project. Therein lies the problem. I suspect many will just shrug and take what AI gives it as "good enough", hence we can all expect a lower quality standard of software going forward once AI advances more.
@@timtanhueco1990 If AGI happens the entire knowledge economy, and as such our entire economic system will cease to function. Along with this US hegemony would probably be challenged as our economy is not set up for production. Not worth worrying about as a dev imo since all white collar workers will be in a bad position and the western world as a whole would be in a precarious position politically. If you legitimately believe that will happen you should go into the trades or a physical job.
Section 174 can cut both ways - on the one hand it’s takes longer to realise tax deductions on engineering payroll, however because these costs are capitalised (goes to the balance sheet) it has the effect of inflating profitability on the P&L, making it easier to show profits and critically for startups - positive contribution margins. For investors, it means reviewing cashflow statements closely is more important than ever, as it’s now possibly to show a P&L profit, while actually being insolvent!
Thank you so much for giving this perspective. I really need some clarity. I am still open to change my domain to other IT sectors like cyber security but I guess I shouldn't just do that because of AI. Do you have any discord or slack groups where we can connect and discuss?
I write VBA for use in Excel. I've used chat GPT to write the repetitive parts. For example, if I need six pieces of code to do the exact same thing two different sections of the workbook, I can get chat gpt to copy the code and change the references. I can't imagine getting it to write the code in the first place. I don't know how it would be able to look at the workbook and understand what the workbook is even doing.
does 174 applies to sock options? like for example we can reduce cash compensation to like 50k and have 900k in stock options? and do stock buybacks for keeping cash flowing into right hands 😄
Mmm... what kind of devs are these? I'm a lead dev and there is practically nothing I do in my day job I could off load to AI and be comfortable using in production.
@zoeherriot There's are lot of "devs" working for small ma and pa companies that basically write VB scripts and Excel macros lol. These people don't work in modern software engineering teams, which you likely do. I'd take their "dev experiences" with a warehouse full of salt.
There is never such a thing as 'senior developer.' Every day is a new day with new sets of challenges and ever-changing customer requirements. No developer is comfortable with all the software know-how. If you are talking about generalist, then AI is doing that job pretty well already.
The idea that near zero interest rate policy caused the surge and fall of investment and thus employment is just not well founded. Japan was had zero interest rate policies for decades, this did not cause excessive borrowing, economic growth, and inflation. The US for many years experienced low inflation with very low interest rates. Today, with high interest rates, we have higher inflation than in previous. The stock market these days are very active. Investors are still pouring money in tech companies. The interest rates during the housing boom were much like around today 4-5%, are we in a housing boom? I don't think so. These things are complex and cannot be reduced to simple cause and effect. Here are a myriad of other factors - government budget deficits, demographics, wars, pandemic, supply shock after effects, oil prices, globalization, AI automation or tech trends, politics, macroeconomy (potential recession), corporate trends, innovation and leadership slump, credit cycle.
I'm with you on this. There's a few more levers at play. Thanks for highlighting those additional levers. I truly wonder how much impact each lever really has at a given moment.
Do u think we were born too early or born too late? I just wish it was 2010~2012 for 4 more decades and the only thing that had advanced was ISP speed. Just wish there's an alien invasion and they can 💩 Gold or extremely valuable materials/resources(all non-toxic) or something, and their tech can be sold if not use for ourselves. It's hard, every land, mine, forest is already owned and monopolized by someone born earlier. Every idea is patented forever by people born earlier. Every idea/concept is made so that if you start one, the older companies will sue you for similarities(how like marvel/DC conceptualize many if not all heroes and sue everyone that's similar to it.)
Not just in coding but in other areas, I find that AI needs supervision. I have played with image and video generators. It has hit on some but it has also missed on many others.
I just wanna point out that your a software engineer worried about AI. Just remember, AI IS software. Learn it, embrace it. Its part of the field and part of the job
But software that makes software is scary. Thats the difference. Not invalidating your main point but to say it’s just like any other tool used to generate code would be an understatement.
Great points! Especially the one about AI doing the heavy lifting. I completely agree that AI is still not there and it be some time before AI could be reliable for larger coding.
Argument on why AI isn't' going to replace all jobs in the long run seems weak to me AI can't understand a large input - This is being improved with each model that comes out. The original GP4 could only go up to 32k tokens, turbo can do 128K tokens, Gemini can apparently do 1million plus (although it seems like its quality is worse). At the same time their ability to pinpoint specific information in a large input is also improving. AI would create small errors, logical or syntax wise - Even current systems are perfectly able to interpret various types of errors and correct their own code. I've had this happen to me numerous times just using the code interpreter with GPT4, it will see a variable isn't defined or a library isn't imported and fix the mistake it made.
A.I does not need to write better code or to replace skilled (human) labor completely. If it can, it certainly will, but a 'good enough' A.I. puts downward pressure on wages. This strategy is very portable / applicable to anything. Consider self-driving cars. A.I. does not need to be a better driver - just good enough to be cheaper to insure. If self-driving cars were cheaper and more profitable (maybe because they aren't liable?) , insurance costs could compel human drivers out of the drivers seat even if the humans were better drivers.
I was just thinking about this tax code problem. One way that a startup could work around it is by giving the software developers different job titles. If their job titles are related to a secondary role (such as sales or operations management), then you wouldn't have to declare them as programmers. There's no law saying that your employees who write code have to be called software developers, is there?
True but the US has job codes that groups or matches with job description. And most tech companies hire immigrants, which has an additional level of scrutiny. Now the job title must match or similar to US Job code AND have a pay commensurate with such title called, prevailing wage determination (PWD)
I agree these are more compelling factors than anything else as to layoffs, I have mixed feelings about section 174 though I want businesses to be more stable and less of these short term gains thinking. And if this forces them to be more strategic about how they spend their resources it might be a good thing. But there will be a lot of pressure to go back to the previous situation because of AI
To your point Section 174 is to starve start-up's and any competition at the behest of the mega-corps. Meta, Alphabet and their ilk, are not affected at all by this. Deloitte and KPMG will make it work. Remember, Bezos got a tax refund - all of it tax code gymnastics. And yes, AI will take over all software jobs with a few folks in the passengers seat. The C-suite at the top don't care if the code is "crappy" or "sexy" as long as it kinda works and they can make it to the next quarter. It's just a matter of time and way closer than you think.
@@catalinagalan It's definitely not caused by a few laws in the US, that is for sure. People like to ignore the simplest answer, which is that right now investments are risky and cash is king. Investors are loss-averse and turn every rock for cash to increase cashflow and valuations today, slashing business development (like VR, etc.) in the process. Pretty much all startups with high burn died almost overnight globally. It's just macroeconomic trends.
Also heard about the grand resignation drive that happened few years back and so employers are now trying to make use of AI as much as possible and keep the headcount low
The risk with AI currently is not whether it can do everything that you can do, it can't yet. But can it help you do your work 50% faster than before? Because if it can and the demand doesn't increase at the same time, then companies can afford to do mass layoffs of say 20% or 10% of their SWE and reduce wages and benefits whenever possible. People think of this in very individualistic terms or in very absolute (all or nothing) terms, but my concern is that even if I'm one of the lucky ones who gets to keep their job, the job might not be as "wonderful" as it has always been, we might be in the brink of facing a new reality.
I don't think the tech will ever get there. The reason is pretty simple too, people who are developing the tech are going down the wholly wrong route. They're the kind of people who are so smart that they're stupid. Most people don't even stop to think what the technology really is, and it's not AI, as the most important aspect for that to be an accurate acronym is missing.
Currently studying so am hopeful I can land a job. But as far as A.I goes 5-10 years is a bit hopeful, you could have a fusion powered super computer and it would likely take 20 years to teach it all it needs to know to think for itself. Since we don't have that it requires a human to tell it every step and correct it a few dozen times before it understands.
Can someone tell me What is the best way to learn coding? I was thinking about studying JavaScript so I can learn about CSS and HTML also, should I study c or c++ ?? I think it’s pretty cool, but I just kinda all over the place learning and watching RU-vid videos and reading books. Thanks y’all have a nice weekend. By the way great content bro have a nice weekend
I've been a software developer / architect for over 30 years now. I've stopped advising people to go into the profession. It's not only the items you mentioned, which are all good points. It's also how developers are treated within a company. We're increasingly not seen as assets that drive profit, but as expenses to be minimized. That's never a situation that will work to an individual's advantage. And I don't see that trend reversing. I also don't see AI as a significant change (yet) but you can bet it's going to be used to intimidate developers and get them to accept poorer terms.
We have a cultural sickness where people that do real work and have real technical skills are devalued as naysayers while folks that can string together emotion-driven corporate babble and swear fealty to the right ideologies can vote themselves $30m bonuses.
It's good to use when you're new to the libs/frameworks/languages or whatever tools you're trying to use. Because the entry level information is widely accessible, the GPT generates usable code. However, the deeper into anything, the more likely AI will completely made it up (Hallucinate)
My Ai rejected all code languages and started to use Assembly Language to code. I asked it to print Hello World, and it told my printer to print Hello World. Now I have a scrap paper. For some reason, it keeps writing libraries. I can't tell what the library does because it's written in an assembly language.
My experience with AI writing code is that it is trained on tutorials with crib notes. Even when I train it with my own code, it reverts back to tutorial style code. There is litteraly no inovation with AI code. The AI code is good for getting a lot of syntax written quickly. Like If I feed it a example data and add new data members, or to write, like Aaron Jack says, a begining template. And then follow up on having the AI do the Google searches for me on technical syntax. And to explain concepts that I have thought out, which I need to apply. The inovation still completely comes from the person writing the code. In my opinon the most valuable aspects of computer programmers is inovation, and software architecture building. As far as the tax codes go. Corprations need to pay their taxes. Everthing they do cannot be an excuse not to pay taxes. They need to be grown ups and responsible, just like everybody else
I just feel quite lost. I'm about to finish high school and was planning to go to a university studying computer science. I know that AI isn't there yet, but I don't think I have the security of knowing that it won't be by the time I get into the job market. Would my skills even be useful in 4 years?
Personally, I think skilled developers who intimately know how to write code will be in far more demand in 3 - 5 years from now compared to today. As I stated in my previous comment to this vid, once AI advances more there's going to be many if not the majority of developers who will just shrug and say "good enough", and run with whatever AI gives it instead if either, putting in the hours with never ending prompts to get the small details ironed out, or getting in there and coding all those small details themselves. This will result in a lower overall quality standard of coding because so many developers will lean on AI too much, and keep doing a copy & paste of whatever AI gives them then say they're done the task / project. Employers, business owners and clients are going to get frustrated and will begin looking high and low for quality developers who actually know how to write code, and not just copy and paste from AI.
AI doesnt do anything on its own, it is just a tool used by people. I would say focus on learning foundational & transferable knowledge and dont fall into the trap of learning some transitory high-level coding frameworks that look shiny at the moment. Start keeping track of job postings in your area, it is vastly different depending on where you live. It will give you a clear idea what is in demand. AI might reduce the number of available jobs by a few percent, since less people will be able to do more (even that is debatable), but IMO foundational computer science knowledge wont ever go out of demand for as long as humanity uses computers.
I agree that it would be better to go into something that can’t easily be replaced, like medicine. Nurses actually make decent wages and have job security because people will always get sick in any economy. Or learn a trade (electronics, electrician, plumbing, carpentry, etc.) so you can get started making money in a lot less time than if you went to college and won’t have many thousands of dollars of debt when you start out.
I don't think Jensen was referring to chatGPT and copilot. software in general is fundamentally changing. not something that chatGPT or copilot or you and me can change. these neural networks infused with formal logic reasoning are changing how software and hardware talk to eachother. missing the point entirely
I tried coding with AI. Its not very good. I couldn't even get it to code a simple function. It codes very logically but it assumes the function recives ideal data. You still need to guide it to account for the "unaccountable". I ended up just using it to make a function franework or check my syntax but its not gonna write programs for you from scratch especially if you're module us supposed to fit with other people's.
The rationale for the amortization is that the benefits of the development happen in future years. The 100,000 revenue in your example would not have come from that developer in this year. It makes sense that the benefits are amortized because of the matching principle. You still get to deduct 100% of the salary, you just do so later.
Honestly I think it's a bit ridiculous that companies could write that off originally and im fine with it being amortized over a few years. The longer timeline for outaourced labor disincentivizes outsourcing and helps protect US workers. Tech in the US has experienced a bit of a bubble, and it might be about time that our tech sectors operated more reasonably burning less money and delivering more with less waste. And I say that as a developer that was laid off and found a new job.
You're absolutely right about AI not getting everything right when it feeds you code to solve a problem. It may get 99% of the code right when you ask it to help you with something but that 1% can be a royal pain in the ass during bug fixing.