An interesting article by Janne Hukkinen in Wired has introduced me to a new concept, peer review conducted by artificial intelligence!
The article argues it is risky business, and I agree. While it can streamline the process and lead to faster publishing, I seriously doubt an algorithm could review a paper to the same quality as a human expert. Nuances would be missed. As the author states it would fail to consider values, “such as ranking two scientifically equal texts on the basis of their social relevance”. Hukkinen also tells a story akin to Hal in 2001: a Space Odyssey, where a rogue program sent “terse letters” without the knowledge of the company.
Clearly there is space for computer assistance, searching for plagiarism and obvious errors for one. It can immediately discount some manuscripts based on particular criteria. But peer review done solely by AI? I am sceptical. Of course, I suppose people are sceptical of every new technological development. Perhaps in the future it will simply be accepted as the norm.
The article poses some interesting questions for the future of publishing and science, and is well worth a read. The possibilities really are endless…
Continuing on from my earlier post about Facial Recognition, let’s talk about machine learning. This is a type of artificial intelligence that lets machines ‘learn’. So just as facial recognition software is able to ‘learn’ by building templates of faces, other machines can learn a vast array of useful information.
For example, a company called Jigsaw has recently announced a new technology to help identify trolls and inappropriate comments on websites (via Engadget). These types of systems are usually given a data bank of typical examples of the content they are learning. Here, the machine (known as ‘Perspective’) was given a huge amount of comments labelled as ‘toxic’ by human reviewers. It can then use these as templates to identify more similar comments. Each time it is correct, or incorrect, it becomes more accurate.
Another example, which is a whole lot of fun, is Google’s ‘Quick, Draw‘ game. Simply press play, and you’ll be given 20 seconds to draw a simple item. These doodles are then saved to “the world’s largest doodle data set” to help with machine learning research in the future. As you draw, you can see the machine recognising your pictures (or not) in comparison to other pictures people have drawn of the same item. Give it a go, it’s very addictive!
Other examples include algorithms that recognise and filter email spam, provide you with ‘you may also like’ suggestions and targeting advertisements at you.
So what does this mean for us humans? Is this the beginning of the end a la Terminator movies? Hopefully not. These systems still need a lot of human guidance, but every new development is truly amazing. And if it simplifies my life and work, I’m all for it!
I stumbled across this fantastic piece of cinema a little while ago and just had to share it here. This short science fiction film entitled Sunspring was written entirely by artificial intelligence (AI).
Basically some really clever guys ‘fed’ this robot heaps of Sci-Fi scripts, the AI (known as Benjamin) analysed patterns within the text to produce an original screenplay by imitating structures and predicting common patterns. It. Is. Amazing.
It even did pretty well in a Sci-Fi London film contest. Before you watch the short film, I definitely recommend reading a bit more detail about its background to truly enjoy the experience (in this great article by ArsTechnica). At least, read the first section of the article which details how the actors managed to put it all together. It adds an extra hilarious and interesting level to watching the short film.
Interestingly, the article also discusses how the script is really a “mirror of our culture” since the AI only analyses existing content and produces the most common patterns into a new screenplay. The detail of how the creators built ‘Benjamin’, and whether he can be considered an ‘author’ are also well worth a read.
And so , without further ado, here is the film – enjoy! 🙂