The State of Artificial Intelligence
July 26, 2018  

Supplied by entelect2013 Administrator from entelect2013
Entelect introduces the "State of Artificial Intelligence" 2018
 

 

 

Artificial Intelligence has really woken up as an industry topic in the last two years, and it’s clear there is still a lot of confusion about what AI really means to us as consumers and businesses. Are we going to lose our jobs or become subjects of a runaway robotic overmind in our lifetimes?! I think a key challenge we face, especially as technology experts in industry right now is to demystify AI for everybody, to separate applications of it into things we can all understand. For example, we’re starting to see that there is a clear difference between an AI that follows rules and an AI that learns. 

 

The rules AI has a strict set framework and leverages formulaic computational strength to win out in an algorithmically intensive battle between a database and a human mind, but it is still very bounded. A well-known example is grand master chess player Garry Kasparov losing to IBM’s deep blue AI. While it’s despairing to know we may never beat Windows at chess again, in many ways these tools should replace menial tasks and jobs for us, and we should be happy for it because we’re then free as a workforce to do things that have more meaning: like planning and interpreting. Businesses should be excited about these, not because it is sexy AI, but because it is efficient. We rejoice as repetitive tasks disappear from our daily work, but as Kasparov himself pointed out, it’s one thing to design a computer to play chess at Grand Master level, but it’s another to call it intelligence in the pure sense. It’s simply throwing computer power at a problem and letting a machine do what it does best.

 

The learning AI is a little more interesting and wild, and we’re really at the early stages of how to apply something like this. Also borne from newly accessible computational power, these AI systems actually teach themselves through observation and data. Some famous examples are Microsoft Tay, which famously became racist on Twitter in less than 24 hours, and Google’s image recognition platform which we’re all training each time we click pictures to ironically prove we’re not a robot online. For businesses looking to apply AI like this, it’s where things become tricky. First recognising that what you need is an AI that learns rather than an AI that repeats tasks with no experimentation, and then when taking the next step to accept that learning AI is still trained by its inputs. In the software industry we have a widely shared philosophy of “garbage in, garbage out” and this still holds true: just refer back to Microsoft’s Tay. This is where Google and Tesla are doing things right, they’re sourcing as much data as possible, with huge variety and volume to train their systems, giving autonomous driving and image recognition the best possible chance of succeeding.

 

Ultimately, as a participant in the global technology industry, we have this responsibility to help our customers and partners cut through the hype, find real applications of what is incredible tech which can make it into production.

- Matthew Butler, General Manager at Entelect


 
PAIA Manual | Privacy Policy | © Copyright by Entelect. All Rights Reserved.