Hello devs!
Please help a fellow dev make a big career decision.
I am a person who is fascinated about AI.
So after working as a gameplay programmer, I have decided to switch my role as a R&D engineer in the same company. I will get to work on cool stuff in the ML and AI domain. But I have got this another job offer for a full stack developer role and the salary is supposed to be three times of my current package. It's great company but the only thing is that they do not have ML and AI in their tech stack. It has been only a year since I graduated, So I wanted to know what would be a good path. To follow what you like or to follow general software development with a great salary hike (which I am sure it would take many years to reach that amount in my current company). Also there are very few companies that offer such a good pay. I want to know that if I go with the salary option, Would it be possible for me to get into the AI domain at a later stage? I would appreciate if you share your experience as well.

  • 5
    I was always fascinated with game development however the salary i get now as a backend dev for an app is 3 times that i would get as a game dev in this country. I never regretted it because in the end of the day you can do what u want in ur spare time. Its not as efficient but getting paid a decent salary is worth it
  • 7
    Wanting to "work on cool stuff in the ML and AI domain" is a very common thought among developers that have no idea what they want.

    You can start by understanding the difference between ML and AI, and then you are likely to understand that there are few to none actual jobs that deal with AI, and not just fart this buzzword.

    And that without years of solid math background you are also unlikely to do anything related to ML.

    Cleaning datasets and integrating ML python packages with no clue as to why they work and how are they implemented, is not ML.

    So my advice, is that you delve deeper into that subject, understand exactly what you want to (maybe you'll discover it's boring and not for you), and then the decision will be much easier.
  • 1
    @NickyBones I know these are buzzwords but I am not talking about them without having any knowledge. I have had done internships and projects in ML and always found the work fascinating. I am also exploring how interesting it is to make AI play a game on its own using reinforcement learning. The thought of having the power to create something that can learn to function on its own blows my mind.
  • 4
    @lightbringer Then you should know what they refer to as "AI playing a game" is not AI...
  • 1
    @NickyBones Except suckerpinch's Learnfun and Playfun series on YouTube. That does have a lot to do with it.
  • 1
    @NickyBones Not necessarily. I mentioned reinforcement learning for a reason.
  • 3
    @lightbringer Necessarily. There is a whole lot of hype around RL, but it is just one paradigm out 3 in ML, and still targets very domain-specific problems.

    To say that a software that can perform a specific task is AI, is a misuse of the word. It's like saying barcode scanners can read, and even faster than people!

    Check out the older definitions and research directions of AI, and see that more than we advanced in this field - we have simply dropped our standard of success.
  • 1
    Here's a recommended read about the subject. It probably explains my views, far better than I am able to

  • 1
    @NickyBones I have said that a software can "learn to" perform a task "on its own". So your argument of barcode scanner is pointless here.
  • 2
    @lightbringer I don't agree with the definition of learning on its own. Formulating everything as MDP and throwing some dynamic programming does not equal independent learning.

    And you didn't address at all the issue with those things being super specific.

    Read the paper, written by people smarter than me, with perspective wider than both of us have, and then voice your objections.
  • 1
    @NickyBones I'm sorry that you don't agree with the definition of learning on its own but that is one of the basic definitions of ML. I would definitely have a look at the paper you shared. I would like to tell you one of the many cases where it is indeed AI that is learning to play the game eg: AlphaGo.

    Here's a nice tutorial playlist on AI playing GTA V by sentdex https://youtube.com/playlist/...
  • 2
    @lightbringer AlphaGo is actually addressed in the paper. I don't see hyper-optimized statistics and costly hardware as learning.
    Definitely not when considering what AI was originally aiming for.
  • 2
    If the salary allows you to live decently there's nothing wrong with the lower wage.
    However be wary that you could still gain enough to retire early and spend the last years of your life doing AI properly.
    @NickyBones AlphaGo is more about ML than AI, strictly speaking. It's not t a revolution but it's a nice piece of software nevertheless. I think the big deal is that AI's and ML's definitions are quite lax.
  • 1
    Depends on what you like to work with.

    A relative of mine is working in ML, and I have to say, he has almost forgotten how to code. That being said, if you enjoy algorithms more than coding, ML and AI it is. Although I agree with @NickyBones , you can just go for the higher salary and work on ML/AI independently.
  • 3
  • 3
    I believe most AI written today except the top 1% is just reuse of libraries and statistics. It's not technically AI at all.

    I would choose the higher salary option if I was in your place.

    Remember, no one can stop you from learning something. If you want to learn AI, go ahead and do that in your spare time. Then you can apply to another company with AI knowledge along with backend experience from this offer.
  • 0
    @NickyBones Am a student and considering to specialize in ML. Is the job prospect not that good (i.e. rare)?
    Also don't really like algorithm, I like building (and breaking) things better... Maybe I should switch to another topics :/
  • 0
    @cho-uc huh ML is algorithms...so if you don't like algorithms I don't see how you're going to like ML.
    There are software developers that specialize in optimization, and participate in writing ML software (coding). But I think the job still requires considerable understanding of the algorithms and very good knowledge of numerical schemes.
  • 0
    @NickyBones AI is one the most misued tech terms out there. Together with blockchain. Even those who should know better, still use it. I gave up arguing the point you are trying to make sometime ago. But when the Singularity will arrive, they will know the difference...
    In any case - You can use ML without a solid base in math, as there is sometimes a need to be able to identify a the best domain - ML algorithem match, and how to construct your dataset correctly, how to deal with input noise, deal with data biases, decide on how to deal with shifting, that is "around" the core ML processing.
  • 0
    Take the money!

    Nice and simple 😂
  • 0
    @magicMirror I don't consider cleaning data or brainlessly using python packages as ML.
    To fit a ML strategy to a domain, you need to know the domain and know a lot about ML - without the necessary background you'll just do a shit job.
  • 1
    @NickyBones very true. However - not many are able to know enough ML internals, *and* domain details to be truely effective in it. Most just nibble at the edge - and using python libs to "build my own NLP AI".
Add Comment