The Limits and Possibilities of Artificial Intelligence
The term Artificial Intelligence is somewhat misleading as it conveys that machines have the power to possess “intelligence”, of the human kind, which is not perfectly accurate. While AI and Machine Learning systems have had immense success in finding patterns when they ingest significant chunks of data, this intelligence is not generalizable. Hence, the pattern recognition is only as good as the data that the machine was fed. That is not to say that applications of AI/ML are not growing, or do not have significant scope, it is only to point out that human intelligence is particularly unique because of our ability to be creative, curious problem-solvers and rapidly switch across domains, and question even the basic principles of a discussion. Such ability is not likely to be seen in machines, and hence the age when they are “sentient beings” is far from approaching.
Why is AI/ML such a big deal?
As Andrew McAfee mentions in his article, “AI, For Real”, AI has led to a significant shift in the way technology is adopted and implemented. The word “coder” is used for a programmer because they explicitly codify, or program a set of processes or algorithms that the computer must obey to produce the desired outcome. Due to this, the possibilities of a computer were limited by human intuition and capability. However, the advent of AI and Machine Learning mean that a programmer does not need to explicitly tell the computer what to do, but instead needs to feed the computer sufficient data and allow it to become better at recognizing patterns as more data comes in.
Applications of AI/ML
Any situation where there is data such that there are input variables (X) and an output (Y), there can be a mapping from the inputs to the outputs. A supervised learning system is merely an algorithm that maps for inputs to outputs. For instance, from historical stock market data, to future stock market data; or from a store’s transaction detail to the output label of whether a transaction is fraudulent. Thus, we see applications of this input to output mapping in speech recognition, trading bots, image tagging, Pharma treatment efficacy, food or movie recommendations etc.
Thus, we can see that any situation where we have data on mapping between inputs and outputs, and we care to predict the output from the input, machine learning and AI can be useful. This means that a number of previously manual tasks can perhaps now be automated, but this also means that the limits of AI are where there ceases to exist high quality data and potential for mappings between inputs and outputs. For this reason, a lot of broad strategic questions will still require human expertise and judgement; predicting the future, when the past is not following a similar pattern, is simply not something that AI systems can do.
Works cited:
https://hbr.org/cover-story/2017/07/the-business-of-artificial-intelligence
https://hbr.org/2017/07/ai-can-be-a-troublesome-teammate
Users who have LIKED this post:
6 comments on “The Limits and Possibilities of Artificial Intelligence”
Comments are closed.
Thank you for this very insightful article on AI technology Adeesh.
There has been an exponential growth in the AI industry this past decade. Though not a perfect representation of human intelligence, AI does simulate a variety of human thought processes. However, as you have stated, AI technology is barren of creativity, inspiration and total autonomy (in regards to its computational processes). The algorithm format to observe patterns do hold its limitations.
Regardless of its limitations, AI technology is prevailing to be seen everywhere in the following decade. Needless to say, there is a massive change in the job industry bound to happen. In a CNBC News report, Kai Fu Lee-founder of Sinovation Ventures-projects AI technology as replacing about 50% of jobs in the next decade. This is a major feat in technology, but also holds problems for the entire labor market. Just as technology has been exponentially growing the past few decades, i.e. Moore’s Law, it seems technology will be transforming our whole economy at a rapid pace as well.
However, there are predictions on new jobs created by AI technology. In fact, there is an MIT study stating 3 categories-Trainers, Explainers, Sustainers-are to be formed for the AI labor industry. It’s quite intriguing though I am not sure this will create enough jobs to fill those being replaced. Moreover, the process of educating these workers shall be a challenge as well…what do you think? Does AI technology have the capacity to change half of the labor market? Also, how should we respond to this scenario if it is bound to happen?
Personally, as shown above, I do believe AI technology will prove a major change in the job market. I am skeptical as to whether it’ll be 50%, but I do vouch for 25% as most of our labor market depends on computers. AI technology could replace the more simple of computer jobs through its pattern-like analysis. Also, if this happens, I would also vouch for a government wide education system to teach workers how to operate with this AI technology.
Please let me know your opinion as I am very interested in this subject matter.
-Jason Vott
Link to CNBC News Article:
http://www.cnbc.com/2017/04/27/kai-fu-lee-robots-will-replace-half-of-all-jobs.html
Link to MIT AI Job Study:
http://sloanreview.mit.edu/article/will-ai-create-as-many-jobs-as-it-eliminates/
Users who have LIKED this comment:
Thank you very much for your article. It is true how you explain how AI works and is at the end just the work of pattern matching and clustering of data. But if you do it extensively, I believe it can replace many jobs. Especially the accounting, Finance branch will be affected by it. From internal sources, I even know that on law-decisions AI products start to emerge and can even replace more complicated products.
Even it is not about that the jobs completely vanish, It is more about the amount of people to do the same job will decrease. AI will never completely e.g. replace accountants; it is just that instead of 10 Accountants maybe just 2 Accountants can do the same job.
And in my opinion, undoubtedly it will affect any branch in the economy and have an impact on it.
Thank you for the thoughts on this widely discussed topic, Adeesh.
I like to bear in mind that many people have the emotional bias of the imagination of job cuttings when confronted with this topic.
I am convinced, that AI and machine learning are radically innovating tools. However, humans have ever since sought and found ways of simplifying tasks, improving productivity in order to ultimately increase overall wealth/ quality of life. Now, AI and machine learning are such improvements as they seek to support human intelligence and work by doing tasks that computers are simply better at. In the long-run I believe these tools to support us significantly to improve our performance and productivity, not replacing it. A focus should be put on how we adapt to this paradigm shift, i.e. how we can learn to utilize the new opportunities brought through these technologies.
Moritz
238a
Thank you Adeesh for your post!
This is a very timely discussion as even Elon Musk and Mark Zuckerberg are debating about the future of AI and it’s impact. I believe the discussion has taken a bit wrong direction between them and here is a good post from Andrew Ng, professor here at Stanford and one of the thought leaders in the field https://venturebeat.com/2017/07/25/ai-expert-worry-more-about-jobs-than-killer-robots/. Some studies indicate that AI would enable significant economic growth in the future https://www.accenture.com/us-en/insight-artificial-intelligence-future-growth but this also requires the labour force to adapt to automation and learn to work with machines and develop new skills for the 4th industrial revolution as discussed in this post from WEF https://www.weforum.org/agenda/2016/01/the-10-skills-you-need-to-thrive-in-the-fourth-industrial-revolution/. Here is a post and some thoughts also from HBR on how AI will impact consulting https://hbr.org/2017/07/ai-may-soon-replace-even-the-most-elite-consultants. Finally, I think Andrew & Erik are great and I warmly recommend reading their latest book Machine Platform Crowd https://www.amazon.com/Machine-Platform-Crowd-Harnessing-Revolution/dp/0393254291
First of all, thank you for this article Adeesh Goel.
I found an article from newsweek that talks about AI taking over the world and they have some very interesting data, for example they say there is a 50 percent chance that AI be able to perform all human tasks better than humans in 45 years, and all human jobs are expected to be automated within the next 120 years, according to a survey of 352 AI researchers who published at either the Conference on Neural Information Processing Systems or the International Conference on Machine Learning in 2015. The survey was conducted by the University of Oxford and Yale University. Also, survey respondents predict that AI will be able to translate languages better than humans by 2024, write high school–level essays by 2026, drive trucks by 2027, work in retail by 2031, write books by 2049 and perform surgery by 2053.
The reality is that machine learning will have a dramatic impact on all aspects of people’s lives, and it’s going to happen sooner than most of us realize. It’s a little frightening, especially if you have a job that involves anything technique-based.
Reference: http://www.newsweek.com/artificial-intelligence-will-take-our-jobs-2060-618259
Some of the major goals that we would like AI to achieve to perfection is Prediction and insight analysis, Natural Language processing – speech recognition, chatting and translation, Image recognition and video understanding, general intelligence and learning, motion and manipulation, creativity and social intelligence. We would like to delegate some of our most time consuming jobs to the computer systems. We want AI to perform real time environmental analysis and reaction in systems such a s self driven cars and crime detection systems. AI systems should be able to analyse symptoms and perform diagnosis and suggest treatment in more common diseases so the doctors can have time to solve more complex problems. The applications of Aı are limitless; we can only get better as the systems improve in performance. We are going into a future where AI will do almost anything that the human can do.