The first part of this blog talks about background, so if you’re keen to read instead about my new chat-bot Blakeley Raise, just skip down a few paragraphs… I’m very excited about Blakeley Raise, and hope you’ll check out the new possibilities. If you can’t wait to give it a go, click here.
So, the background… I had the great pleasure of going to the technical day of the Microsoft London Future Decoded conference last week. It was packed with all kinds of interesting stuff – far too much to take in in the course of a single day, in fact. There were cool presentations of 3d technology – the new Hololens device, enhanced ways to visualise 3d objects within a computer, and how 3d printing is shaking up some parts of the manufacturing industry. And lots of other stuff.
But it all threatened to be a bit overwhelming, so I kept my focus quite narrow and stayed mostly with the AI stream of presentations. Top level summary: Slate (in Far from the Spaceports and Timing) has no need to worry about the competition just yet, but there is some really interesting work going on. It will take a lot of generations for Slate to emerge! But the work that is being done is genuinely exciting, and a mixture of faster hardware, reliable communications, and good programming practice means that some tasks are now trickling into general everyday use.
One speaker used the phrase “slices of intelligence” to capture this, recognising that real intelligence involves not only a capacity to learn tasks and communicate visually and in words, but also to reflect on success and failure, set new challenges and move into new environments, interact with others, be aware of moral and ethical dimensions of an action, and so on. We are a very long way from producing artificial intelligence which can do most of that.
But within particular slices lots of progress has been made. Natural language parsing is now tolerably good rather than being merely laughable. Face recognition, including both identity and emotion, is reasonably accurate – though the site http://how-old.net/ produces such a vast range of potential ages from different pictures of the same person that one can be both flattered and disappointed very quickly (give it a try and you will soon find the limitations of the art at present). On a philanthropic note, image recognition software has been used to provide blind people with a commentary of interesting things in their immediate neighbourhood: see the YouTube snip at the end of this blog.
Here’s the bit about Blakeley Raise… For those of us who develop our own software, it is an exciting time. It is extremely easy now to develop a small program called a chat-bot which can be incorporated not just into web pages, but also message applications like Skype, Facebook Messenger, and a host of others. So inspired by all this I have started developing Blakeley Raise, a bot who is designed to introduce potential readers to my books. You can think of Blakeley Raise as a great-great-ancestor of Slate herself, if you like, though I don’t think Slate will be feeling anxious about the competition for a long time yet.
But one of the great things about these bots is that they can be endlessly reconfigured and upgraded. Right now, Blakeley Raise just works by recognising keywords and responding accordingly. Type in “Tell me about Timing” – or another sentence containing the word “Timing” and you’ll get some information about that book. To find out more, navigate your browser to http://www.kephrath.com/trial/BlakeleyRaise.aspx and see what happens. All being well – meaning if I can solve a few technical problems – Blakeley Raise will soon appear on other distribution channels as well. (For those who remember the episode where a Microsoft bot quickly learned how to repeat racist and other inflammatory material, don’t worry – Blakeley Raise does not learn like that)
Finally, here’s a video of one of the more philanthropic spinoffs from Microsoft’s enthusiasm about AI in practical use…