What should I study I wish to work as an artificial intelligence researcher?
What to study depends on what you;re interested in and what your abilities are. AI is a large enough field that you will eventually specialize in some aspect of it, some problem. Right now you likely don't know yet what will be your path. So, at this point you need to do a general survey of AI topics and get a feel for the whole field. Then you'll end up finding a corner to work in.Some people may study human psychology and cognition, and then try to make AIs that perform the same functions. Other people work on things such as computer vision, how to model the world, recognize objects (snd ask what IS an object, because a thumbtack is easy enough to recognize, but how do you recognize a cloud? or a mountain?). Other people start with philosophy and ask what is knowledge, and then try to see how a computing system can handle knowledge.So you can see that what you study will have to depend on where you are going, and to figure out where, you first have to get the broad overview.
What's your opinion about artificial intelligence?
Wow, that's a really broad question. So let me break it down to more specific questions.When do you think we will have artificial general intelligence?Well Ray Kurzweil made a survey for that and if I don't remember it wrong the most popular answer was about 50 years or so. Well, so far if there's only one thing we have learned about predictions about technology is that they never fail to fail. It always happens in a shorter time than we expected. So I'll say 30 years max.What is the best way to achieve AGI?Well if we knew, there would already be one. There are of course limiting factors other than our ignorence, but I believe there's always a more efficient and simple way to solve problems, and we have not found one for artificial intelligence.How are we using AI algorithms right now?Well, you probably use Google. It uses machine learning algorithms to rank pages. Siri for example uses natural language processing algorithms to understand what you said. It's all around us, watching and learning.Can I develop an AI to do …?Well you probably can by using libraries, which would not require you to know what happens behind the functions such as tensorflow or SciKit Learn. They're mostly for python since it is a pretty easy language to learn for everyone.Ok I want to develop, what now?Try Udacity's machine learning course by Sebastian Thrun. I found it pretty good and easy to learn as he explains both the usage and what happens in the background. If you want to go more theoretical try Andrew Ng's course on Coursera. It's a great course but requires calculus knowledge.If you have any more specific questions, let me know :)
What is it like to take 6.034 (Artificial Intelligence) at MIT?
When Patrick Winston taught it, I thought it was one of the best courses in the major. It was a survey of major methods in AI with some coverage of success cases. I liked the programming assignments, which were in LISP. It gave me a good foundation for later work in robotics and AI.Back in the day they had a bit of a bias (no pun intended) against neural networks. I thought that was a weakness. Neural nets turned out to be the most important method I used in my later work. Not sure if they have revised the curriculum.
What is your biggest fear regarding artificial intelligence taking over human decision making?
Steven Wu is spot-on, and Jason Li has a good point. The idea of the tech singularity is nonsense, and as for human extinction, what makes anyone think we need AI to do that? In The Terminator, SkyNet kills most of humanity with nuclear weapons, of which the US has roughly half a gigaton worth, and climate change is happening and killing people already.Grace Hopper famously said that the most dangerous phrase is “We’ve always done it that way.” The current ML simply encode our decision-making processes. Cathy O’Neil’s wonderful (and frightening) book Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy effectively shows the consequences, but we don’t need to look to ML to see the outcome. Your financial life (often including jobs!) revolves around one number, your credit score. It’s notionally a measure of risk, and we know the components, but we don’t know the full formula. It’s problematic because, while it rewards good behavior (a good thing), it also rewards wealth and access to wealth, and it rewards knowledge of and adherence to societal biases. ML just (potentially) weaponizes our biases.I would two things.First, it’s very difficult to unroll a neural network model and understand how it is making its decision; we can spend the resources when we see a big, clear error (e.g., a self-driving car driving over a puppy), but can we determine why you weren’t admitted to a college that used an ML admissions system?Second, we’ve had several AI winters - Wikipedia, and if current ML doesn’t prove to be the solution to all of life’s problems (spoiler: it’s not), we’re likely to have another.
What are your opinions on artificial intelligence taking jobs?
Sometimes I feel like I am the only AI sceptic on this site.Oh well,my predictions are:Large scale statistical methods will result in some jobs being automatized but it’s not going to be as fast and as drastic as people think it is going to be.There are likely some model limits on the form of distributions that current models can learn/ the number of samples needed to learn these distributions/complexity problems with training for non-trivial input structure that would limit current models applicability.Model interpretability (and training associated instability) problems would stop current model adoption in medicine (it will still happen to some degree but will take years)Do I find the field interesting ? Sure: I would not be working in the field if I did not. Do I think the current hype is justified ? No, I don’t, would like to be proven wrong though.
What are some good tips to study Artificial Intelligence?
6 Easy Steps To Get Started Learning Artificial IntelligenceArtificial intelligence (AI) is a sub-division of software engineering. The main goal is to enable a smart PC/Cell phone to perform exercises that are normally done by people. To start with said back in the 50s in the paper “Computing Machinery and Intelligence”, composed by mathematician Alan Turing, AI is currently an exceptionally well known field, and we have propelled innovation to “fault” for that.. This article is about learning Artificial Intelligence and we will give you a comprehensive guide that you can use as a starting point towards learning artificial intelligence.Where you start depends on what you already know.Below you’ll find a list of resources to learn and practice and how to get started in Artificial Intelligence in 6 Easy steps:STEP 1.) Learn Python & SQLPython is what many prefer to start with because its libraries are much better suited to Machine Learning.STEP 2.) Learn Machine Learning from my listed courses here Top 10 Artificial Intelligence & Machine Learning Courses that will help you turn into the following ML master Google or Apple employs.STEP 3.) Learn basics of probability theory, statistics, Data science and some computational mathematics.STEP 4.) I have listed here some of my favourite free machine learning/Data science ebooks from where you can download and kick start Machine Learning Basics/Statistics for developers to become good at building AI systems quickly.STEP 5.) Practice few exercises on Scikit from website:STEP 6.) Practice practice on your own, step by step you will slowly become AI programmer.I have listed here free open source AI tools which you can use to build your solutionsOnce all these 6 steps are done then you can have a glance at these interview questions on AI and start giving interviews if you want to start career in AI/ML. Good Luck!
Will artificial intelligence kill social media manager jobs?
Along with signaling disruption, a recent survey of 1,000 business and IT leaders commissioned by Infosys which finds AI -- as we know it today -- has moved beyond the experimentation stage, and is delivering real benefits. The vast majority of the executives surveyed, 86 percent, say their organizations surveyed have "middle" or "late-stage" AI deployments, and view AI as a major facilitator of future business operations. In addition, 73 percent agreed or strongly agreed that their AI deployments have already transformed the way they do business, and 90 percent of C-level executives reported measurable benefits from AI within their organizations.Believe it or not, AI is more than just automation. While a majority of organizations in the survey, 66 percent, start off using AI to automate routine or inefficient processes, it becomes a factor in innovation and differentiation as time goes on and experience is gained. For example, 80 percent of IT decision makers at organizations in later stages of AI deployment reported that they are using AI to augment existing solutions, or build new business-critical solutions and services to optimize insights and the consumer experience. The same percentage of C-level executives said their future business strategy "will be informed through opportunities made available with AI technology." Another 42 percent also expect significant impact in research and development in the next five years.
Factor analysis and General Intelligence (g). what exactly is it?
Factor analysis is a procedure that looks at lots of information from a large survey (for example) and narrows it down to the few important ideas by relating the questions statistically. So, there might be 300 questions on a personality survey, but FA can narrow down the core "factors" to 5 (Personality traits). G is the thing we think underlies all the different types of intelligence questions, styles, and behaviors. FA is used to find G. It's confusing because we (psychologists) don't have that good an understanding of what it is ourselves.
How and where do I publish a research paper about artificial intelligence?
Some of the top publications in AI are below. Given as Publication Name --- h5-index --- h5-medianExpert Systems with Applications --- 89 --- 118 Journal of Machine Learning Research --- 73 --- 122 International Conference on Machine Learning (ICML) ICML Lille --- 69 --- 103 Neural Information Processing Systems (NIPS) NIPS : NIPS --- 66 --- 94 IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics --- 58 --- 92 Applied Soft Computing --- 57 --- 92 IEEE Transactions on Fuzzy Systems --- 56 --- 87 IEEE Transactions on Neural Networks --- 52 --- 90 arXiv Learning (cs.LG) http://arxiv.org/list/cs.LG/recent --- 50 --- 78 International Joint Conferences on Artificial Intelligence (IJCAI) --- 46 --- 63 Neurocomputing --- 45 --- 51 Artificial Intelligence --- 43 --- 70 Neural Networks --- 43 --- 68 arXiv Machine Learning (Freenom) Machine Learning --- 40 --- 68Engineering Applications of Artificial Intelligence --- 39 --- 57 MIT Press Journals - Neural Computation --- 38 --- 51 Machine Learning --- 37 --- 60 AAAI Conference on Artificial Intelligence --- 37 --- 50 Robotics and Autonomous Systems --- 37 --- 46 Journal of Artificial Intelligence Research (JAIR) --- 36 --- 66 Source: Google Scholar MetricsFor publishing research papers:Firstly and most importantly, have/find an interest area where you want to carry out research. After this, follow the steps as below:1) Find a research problem: Do an extensive literature survey and find a research problem which interests you.2) Work on it, get some good results: Work on the research problem and obtain good results. By good results, I mean something which improves the current state-of-the art. 3) Find an appropriate journal/conference: Select a journal/conference based on the results and the impact factor/reach of the journal/conference. You can use the Google metrics also to select a journal/conference. Read its scope. Make sure that they publish works in the same or similar area. If you feel that the results will have significant impact, target a high impact factor journal/conference from Elsevier, IEEE,Springer, Wiley etc. 4) Draft the manuscript and communicate: Every journal/conference has its own specifications. After selection prepare the manuscript according to the guidelines. 5) Wait for the reviewer comments: The manuscript will get accepted and then published once it is acceptable by the reviewers.Happy Researching and Publishing!!!