Turbotodd

Ruminations on tech, the digital media, and some golf thrown in for good measure.

Posts Tagged ‘artificialintelligence

Bot to Bot

leave a comment »

Facebook’s been in the news a fair amount this week.

Pivotal Research lowered its rating on Facebook to “sell” from “hold,” according to a report from CNBC, explaining it is “facing digital ad saturation risk as large companies are ‘scrutizing’ their marketing budgets.”

This despite the fact that Facebook has been one of the best-performing large-cap stocks in the market, growing nearly 50 percent year to date.

Earlier today, Fortune reported that Facebook is amping up its artificial intelligence capabilities, buying Ozlo, a small bot specialist based in Palo Alto.

Ozlo focuses on “conversational” bots that talk to users, and most of the company’s employees will join Facebook’s Messenger team.

But the story that really seemed to grab the Facebook headlines this week was the one that indicated two of its bots, instead of just talking to humans, were talking to one another and in a language that the chatbots “invented.”

Before you go all “Westworld” on me, let’s separate the fact from the fiction.

In an account from Karissa Bell at Mashable, Bell provided some much needed background to stifle the hype and get to the actual innovation. Bell wrote that “Facebook’s AI researchers published a paper back in June, detailing their efforts to teach chatbots to negotiate like humans. Their intention was to train the bots not just to imitate human interactions, but to actually act like humans.”

Which humans, we’re not yet sure of. The Mooch? Kim Kardashian? Kid Rock (Soon to be Senator Rock, to you!)

Unclear.

But Bell’s observation was that the narrative wasn’t just about the chats coming up with their own language, but instead this: That not only did the bots learn to act like humans, actual humans were apparently unable to discern the difference between bots and humans.

Where the bot chatter went off the rails was in their use of the English language, the grammar and syntax rules for which the bots were not instructed to use. Hence, some of the shortcut phrases like “I can can I I everything else.”

In the meantime, Elon Musk has cried AI Chicken Little once again, suggesting all this neural networking could be the end of humankind once and for all and that Zuck doesn’t “fully understand” the potential danger posed by AI.

The truth probably rests somewhere in the vast middle ground between the two, a truth I imagine the bots are having a good chuckle over as they create the new digital Esperanto they’ll need to take over the world.

Written by turbotodd

August 1, 2017 at 10:59 am

%d bloggers like this: