Facebook – The dunce.

Reply to this post

AI is emerging as a major weakness for Facebook. 

• There are signs that Facebook is really struggling with automation as it is clear that its algorithms are inferior to those used by Google, Baidu, Yandex, Microsoft, Apple and almost everybody else.
• This will become a major problem as I am convinced that differentiation in digital ecosystems over the next 10 years is likely to be largely determined by AI driving how intuitive, rich and useful digital life services become.
• With 1.6bn active users, Facebook simply cannot afford to do very much manually, meaning that how well its AI develops will determine its ability to challenge Google and take share of advertising revenues.
• Every indication I have seen points to the fact that as soon as Facebook removes people from the equation, things go badly wrong.
  • First: bots. Facebook’s messenger system is a great system for person to person and business to client communication but the automated bots are so stupid as to be effectively useless.
  • Early interactions with Facebook M and the bots that have been launched (see here) show machines that are far too dim to be of any real use.
  • If I was a company looking to build a bot to interact with my customers, I would not be using Facebook’s.
  • I would instead be looking at Google, Baidu or Yandex as a foundation for my machine intelligence.
  • Second: news curation. Testing by the Washington Post found that the removal of humans from the news curation process led to fake news being displayed as trending.
  • In the weeks following the move to automation, Facebook trended old news, fake news and conspiracy theories and was also found to be slow when it came to picking up the real news.
  • This can be hugely problematic as enough people use Facebook for news that bias is becoming a big issue.
  • Facebook needs to be as neutral as it can and while its AI throws out spurious stories and ignores others, it can easily be accused of bias when all is really happening is bad algorithms.
  • It is these errors and unintentional bias that the human curation weeds out as well as being able to quickly tell what is important and what is not.
  • Put simply, this is another example of how Facebook’s AI is not up to the task of automation meaning that an awful lot more work needs to be done.
• The net result of these problems is that Facebook’s services will fall behind those of its competitors unless it can get its AI up to scratch and do so quickly.
• This will be very difficult as good AI takes a long time to train meaning that it Facebook will need to acquire both talent and algorithms rather than create them from scratch.
• In terms of the short-term outlook, this is not a problem as growth is still being underpinned by the increasing dominance of digital in users’ daily lives.
• Over the next 2 years or so, this will slow markedly meaning that AI will have to take up the slack in order to keep growth going.
• This is one reason why I remain cautious on Facebook in the short term as I think consensus is assuming success in areas where Facebook has huge amounts of ground to make up.
• Tencent, Microsoft and Baidu look like better places to be in the short term although I am keeping an eye on Facebook for the long term.