Moving into AI without a clear game plan

Sambhavi's profile thumbnail
Hi Erika,From my experience of self-learning in all the above subjects, here is what I would suggest, take a top down approach, else it is extremely easy to get lost as you will be swamped by different branches and different materials and courses out there.When I say top-down approach, what I mean is, first identify the use case you want to solve, or the solution you want to build, or write down the problem statement. That itself will take some time to formalize. Then move on to the how part. At that time start exploring how best you can solve it using AI, is it a NLP problem or a Computer vision problem or a data problem or something else. Based on that, dive further deep to find the right set of resources.Another point that really helps is to learn by doing, always have a mix of coding and theory, they should go hand in hand, else it's quite tough to keep up the motivation factor. There is no instant gratification possible in this subject as it's so vast but if we pace our learning with a goal (use cases at each point) then we can see some output on and off and that will keep us going.Would not want to overwhelm you with resources and links at this point in time. Hope my reply gives you some direction.
JustErika's profile thumbnail
Hi @Sambhavi, Thank you for taking the time to pen this reply out for me. That seems sound to me. Thank you for the input. I'll definitely begin first with outlining the specifics of the problems I want to work on. Very Much Appreciated, Erika
maddogS's profile thumbnail
Let me answer your questions in the reverse order : )3. This is not only a good idea, it's ABSOLUTELY critical. In the first 1-2 stages of the hiring process, projects are how we filter and also frequently used as a basis for interview questions... we find it's not really a great source of signal if I ask you about a field of ML you never worked on... but it's absolutely fair for me to ask you questions about your projects and use that to gauge fit, how you think about a problem and if we want to proceed with next steps (a take home project, "onsite" interviews etc...)... for some details about how interviews work and what kinds of questions you might get around your project portfolio: https://www.themuse.com/advice/data-science-interview-questions-answers2. If you are just starting out I would recommend learning python. The industry has decided to pick python as the "lingua franca" of data and ML. Which means more ML / AI jobs will require python and there are a lot more resources to help you. (Just knowing python is not enough but not knowing python is definitely a deal breaker). If you would like to focus more on performance and engineering I would recommend C++ as a 2nd language1. going all the way back to answer for 3. what I recommend is, pick a data set or a social problem you find interesting... learn as you go to answer your own questions, let your curiosity drive. This will not only tell you what exactly you need to learn next...it'll also 2 birds 1 stone for your portfolio (don't forget to present it nicely in your github readme and even better a medium blog : )I wish you fun and success! P.S.Here is a mis-mash slide deck I use for some past talks on AI topics... The first section is about an overview of different options in the field of data / AI, what the job typically require and some resources on how to get going... which might be useful for you.(if you go past that... note the tensorflow code in the slides are based on TF1.)https://docs.google.com/presentation/d/1Xi4IrOkM7hJFy98WzSQ8LWCm8rrUKpi7FiRXrrqNOOI/edit?usp=sharing
JustErika's profile thumbnail
Hi @maddogS,The holidays interrupted things a bit for me so I hope you don't mind the very delayed thanks. I've already begun reviewing the materials you included in your response. I'm particularly encouraged by slide #4 in your presentation. :> Mind out if I reach out if I have any other thoughts/questions? I promise not to be a pest! Best,E
maddogS's profile thumbnail
Happy to help... please feel free to PM me. (I'm also a bit swamped at the moment, so response times might vary...)Also happy to answer in a public forum... in case others could benefit from your questions... : )
maggiewolff's profile thumbnail
Hi Erika, I'm halfway through a masters degree in data science and currently work in analytics, so while I don't have experience in AI, I have wrangled with some of these same questions. For programming, I would start with SQL and then move on to Python. SQL is used by so many companies to query their data, so even though it isn't a "fancy" language like Python, it is necessary because if you can't get the correct data then your analysis and modelling will not accomplish your goals. SQL is also pretty easy to learn, relatively speaking. After that, Python is very commonly used by data science and machine learning teams. There are a lot of "Python or R?" debates in various data science communities. They are pretty similar, so if you know one, you can easily learn the other. Since Python is more popular, it's probably better to start there. For math, start with basic statistics (the stuff you need for hypothesis testing), and also linear algebra.
JustErika's profile thumbnail
Hi @maggiewolff, Thanks for penning this out. I hope you don't mind the delayed thank you! I'm learning SQL at the moment, and I'll be following that up with a brush up in Python. Will do with starting off with Stat. Best, E