In the week that Apple started to finally open its doors and discuss Artificial Intelligence, some other interesting stories started to appear. More on those later.
First, let’s look at Apple. Lately, they seem to have fallen down in analysts and people’s expectations. By generally not releasing the best stuff, it seems that others are catching up with Apple’s early progress and are now overtaking them. We could have assumed this was the case with Artificial Intelligence as well until this week.
Often hidden behind a veil of secrecy, Apple seems to have realised that if they want to attract the best talent in the AI world, one that is currently paying huge salaries and working with some of the most amazing companies, they need to be more transparent about what they are working on – something very much against the philosophy of Steve Jobs in the past.
As the conclusion of the article states: “If Apple wants to stay ahead of its competition, it has to finally start giving away its secrets”.“If Apple wants to stay ahead of its competition, it has to finally start giving away its secrets”.Click To Tweet
But this segways nicely into another article this week, one that is slightly more philosophical in approach. We may be attracting the very best talent into the sector but it is the bias of these incredible people that is perhaps defining the industry and what it is creating.
In an article for Techcruch, entitled 5 unexpected sources of bias in Artificial Intelligence, the author covers the misgivings of how we build an AI and the bias it creates; in summation “machine bias is, in fact, human bias”.“Machine bias is, in fact, human bias”. Click To Tweet
To precis, they break the bias in AI down to these 5 sources:
- Data-Driven Bias: For any system that learns, the output is determined by the data it receives
- Bias through Interaction: think of Microsoft’s Tay, the Twitter Chatbot they had to turn off after 24hrs as it had turned aggressively racist
- Emergent Bias: think of the bias “bubbles” of our socially reinforced view of the world through our Facebook newsfeed. Does this mean we are seeing what we should or just what FB thinks we should because of our friends?
- Similarity Bias: Sometimes bias is simply the product of systems doing what they were designed to do. Here they state Google News as the example. It does exactly what it is supposed to do and nothing more
- Conflicting Goal Bias: where systems that are designed for very specific business purposes end up having biases that are real but completely unforeseen. For example, stereotypical bias. There is an interesting case study in the article so work taking a look.
So there you have it; the industry is paying for and charging after the very best of AI talent to help create the latest and best AI. However, we need to be aware that every AI built will have an innate bias, depending on said build or simply on how we are interacting with it.
Big challenges for big minds. A fascinating industry to understand more fully as it matures.