My Weekly Journal
First Journal Entry!
2/10/18
Welcome to my Journal! I am now going to upload a new journal entry every Saturday morning, and each one will be about anything interesting that happens. Hope you enjoy!
TEDxBerkeley 2018
2/17/18
I just came home from a tiring day at TEDxBerkeley 2018! I’m not complaining or anything, the whole experience was really fun. To add to that, I was a guest speaker, and I gave a talk on the history of AI. Since I’m really tired, I’ll just have a link to my powerpoint of the talk I gave. Hope you enjoy!
An argument about AI and warfare.
2/24/18
Many people I know have recently told me about some really opposing views about how AI should be used in warfare. My friend Ben said that AI should be used because it’ll save so many lives, but my other friend Harry said that AI shouldn’t be used in war because it’s not ready yet. I decided to do some research to find out myself.
From what I’ve found out, AI really shouldn’t be used in war, as Harry said. First of all, as my fellow BAIR researcher Michael Luo said, the best of AI right now can only learn to recognize cat pictures and play simple games like Atari Breakout (Luo). Bring something like that to the front lines, and who knows what it’s going to do.
Also, if you do bring AI to the battlefield and it does something bad, you’re not going to be able to explain what happened and stop that from happening because of the way AI learns. In commercial AI that doesn’t learn, decisions are made through an algorithm, which is like a math equation, and it doesn’t change until you change it. With an algorithm, you can trace what it did to why it did it. However, with learning AI, the most powerful of all AI, it approximates an algorithm, meaning that there aren’t concrete steps to how it got to a conclusion (Luo). Approximating algorithms makes AI more flexible and powerful, but if it does something wrong or bad, you will never know why it did that, and there isn’t a way to prevent that from happening again, a problem on the battlefield.
I Got Interviewed!
3/3/18
A lot happened this week, but what I think was most interesting was that I got interviewed! Chanelle Baldwin, a student from Albany Middle School interviewed me. I talked about the strength of AI and what that means for the question of having robot overlords. Here is a link to the transcript of the interview. Hope you enjoy!
More about war.
3/10/18
In last week’s blog, I talked about how robots weren’t ready for war. A couple days ago, I found these two GIFs above that paint a very good picture of what I mean. The GIF on the left makes AI look at least a little nimble for being able to tie a knot, but the GIF on the right immediately shows otherwise. I don’t think anyone would want a robot like in the right GIF on the front lines of a war.
New Essay: The Future of AI
3/17/18
There's no time stamp on my journals, just a date stamp, so I'll tell you now that I am writing this at 11:30 PM because I just finished editing an essay I wrote and I just sent it in. Here's a link to it, since I really need some sleep.
Cyberwarfare.
3/24/18
I was out having a coffee with my fellow researcher Michael that I’ve talked about in previous blogs when he reminded me that we could use AI in cyberwarfare. It blew my mind, because, after thinking so much about using AI in war, I completely forgot about cyberwarfare, something that is so close to AI.
Michael told me that something AI would be very good at would be detecting a hacking presence, and learning ways to defend systems from attack, while at the same time learn to hack enemy computers with an efficiency that probably already outpaces humans (Luo). A good way to control the AI is to force it to wait for approval before hacking and to only be able to hack approved targets. AI could also suggest which targets would be easier to hack or have more of an impact. In this way, AI should definitely be used in war, as long as it’s cyberwar.
TIME Article!
3/31/18
A while back, I sent in an article about why AI wouldn’t affect our jobs to TIME magazine, and they just replied saying that they’d publish it in the next issue! I’ve been pretty busy lately, so I’ll just give you a link to my article for today’s blog.
