A response to One Hundred Year Study on Artificial Intelligence

The 2016 report of One Hundred Year Study on Artificial Intelligence covers in detail the situation Aritificial Intelligence faces and its status currently, and future possibility on policies and itself. In the fields where AI is relatively hot it addresses, I personally find several points that worth more discussion.

First of all, it says that the unexpected interests human have put into the entertainments based on AI technologies have caused a drop in interpersonal connection.(A100, 2016, p.40) I do not think this is necessarily true. According to what they have written in previous texts, entertainments based on AI actually include social network platforms and instant message services, which are obviously intended to strengthen the bonds between individuals. For example, Through Facebook, one can keep close to their family and friends even from a far distance (Koifman, 2013). Social networks provides a mean for people to stay in touch with the person they want to interact, even under the condition when traditional ways to contact like mails or in person cannot be applied. As a result, more interaction should strengthen the bonds which connect people together, just as meeting face-to-face does. What's more, such tools update really frequently with new way to convey feelings which are even more interesting. If I remember, Google's Allo, the IM software allow users to adjust the size of their bubble, which give additional info like how strong one's feeling is at the time besides the literal meaning. Doing so the user can have a great lot of fun even they are not meeting the actually person they are sending message to. Never forget to mention that nowadays, these AI based tools mostly have video chat feature which directly allow user to interact with each other. Thus, I do not see the problem on decreasing interaction in humanity with more AI entertainment.

In addition, the report mentioned in page 43, that AI can be biased for the point of view of its designer and the data it relies on. An example the report comes up with is about the difficulty speech recognition techonology meet while the users are women or have heavy accent. Personally, I think this problem does exists. Yet, it is relatively easy to solve, and is already done on several products we widely use now. Please check the ever repeating videos on Youtube about tests on voice assistants. A whole lot of them are about the test under different context of accent. There is one in particular tests on Google Assistant, Apple Siri, and Amazon Echo, in which Japanese, French, and even Russian "flavored" English are applied in the experiment (WIRED, 2017). And as expected, the voice assistants can all handle such accent heavy pieces of English to a certain extent, in which the Google Assistant have the best overall performance. I believe it has something to do with the dedicated work the developers have put into the voice sampling phase. They get large set of samples, including woman and accent-heavy ones, and the AI learns well.

Finally, on page 43, the report also address the problem that the economic fruits created by AI's power is likely to be shared evenly, mostly to the hands of the scarce of expertise. However, such issues should be mitigated by fast widening access to AI really quick. In recent years, there are more and more emerging online courses on AI, such as, you know, the Coursera one. Though most of them are introductory level, it is still important that wider access is provided to the public, if they want to know more about AI. More than that, the problem on the cost for AI-ready hardware is now partially solved by cloud based computing service, like those provided by Google, Microsoft, and Amazon. And the hardware itself has become even cheaper for ordinary customer level: Epyc processor lines on zen architecture from AMD, much cheaper and far more powerful than its Intel counterpart. And most importantly, machine learning specialized hardware from companies like Google and Nvidia. All of these should empower the AI to spread its user base wider than what it is currently in the future.

Overall, the report is a good read and let me reflect a lot. But I expect studying the search section on our textbook on the weekend of the first week, as some preparation to the incoming challenging lectures.


Reference List (sequence they appear in this paper):

Peter Stone, Rodney Brooks, Erik Brynjolfsson, Ryan Calo, Oren Etzioni, Greg Hager, Julia Hirschberg, Shivaram Kalyanakrishnan, Ece Kamar, Sarit Kraus, Kevin Leyton-Brown, David Parkes, William Press, AnnaLee Saxenian, Julie Shah, Milind Tambe, and Astro Teller.  "Artificial Intelligence and Life in 2030." One Hundred Year Study on Artificial Intelligence: Report of the 2015-2016 Study Panel, Stanford University, Stanford, CA,  September 2016. Doc: http://ai100.stanford.edu/2016-report. Accessed:  September 6, 2016.

Koifman, N.(2013). Can Social Media Actually Benefit Relationships? Retrieved from: http://www.huffingtonpost.ca/natasha-koifman/social-media-and-relationships_b_4115588.html

WIRED. (2017). 8 People Test Their Accents on Siri, Echo and Google Home | WIRED. Retrieved from: https://www.youtube.com/watch?v=gNx0huL9qsQ&t=263s
