Can AI Save Social Media?
For a good time, try watching the weekly “Prime Minister’s Questions,” posted by the British Parliament. It seems that every Wednesday, the Prime Minister must face the House of Commons and answer whatever questions are asked of him by often hostile members, accompanied by (at times) catcalls and cheering, booing and eye-rolls. The whole thing is posted on the parliament.uk website. Fun.
To say that British politicians are a bit more likely to engage in unvarnished truth-seeking and truth-telling is an understatement. It’s a more lively system over there.
So it was interesting when a contingent of the U.K. Parliament came to Washington in early February to question some executives of Twitter, Google and Facebook about the quality of the news and opinions hosted on their platforms.
Some of the questions asked were: “Why has your self-regulation so demonstrably failed, and how many chances do you need?” “How can your system be described as anything other than inadequate?” and “Is this too much for you?” Not the usual fawning acquiescence that the tech giants expect.
The Brits were here examining Russian-funded interference in the voting on Brexit, and looking to understand how false news and deliberate disinformation affect British elections and society.
Other British inquiries have been launched into why it takes so long for Twitter to take down defamatory and often anti-Semitic posts, even when flagged immediately and repeatedly.
Some legislation no doubt will come out of this, much to the tech giants’ dismay. They haven’t taken this seriously enough in the beginning, and it has overtaken them; and they need to respond to the international market.
Twitter actually lost U.S. users last quarter, down by a million, but the overall international membership was up 4%. This led to it making a profit for the first time ever, led by an increase in ad revenue. But many overseas governments are more restrictive than the free-wheeling U.S.
So the tech giants are turning to their always solution: more tech.
More Tech, Please
Facebook recently shuffled its artificial-intelligence (AI) management, moving the head hired in 2013, Yann LeCun, to a more limited role and bringing in a leader from IBM’s Watson AI platform, Jérôme Pesenti. He will have to deal with the backlash to both election-meddling and impacts of social media on mental health. No easy task.
The prior head certainly was no slacker. The facial recognition qualities of Facebook are often eerie. Using the “tags” that people put on photos, they have developed an astounding database of people’s features, which then can be used to select what shows up in your news feed or the “people you may know” feature, which aims to entice you into friending more people, or at least keeps you engaged.
AI also has been significant in targeting advertisements and helping translations.
The next challenge is putting a muzzle on inappropriate content. While it is easy to scan posts for things like “you dirty %*#@,” the problem is deciding if it’s a hate post or a birthday greeting from that particular relative you hesitate to invite to Thanksgiving (or both). The solution no doubt will depend on better recognition of the “bots” that have been used quite successfully by the Russians to spread division and animosity.
This won’t stop that crazy relative from passing things on, however might kill some of the “trending” features that helped make fake content so successful.
Twitter is another whole ball of wax. It never has had any restrictions on bots, something that allowed fake hashtags and nasty content to spread rapidly and effectively. A good reason to avoid it entirely. So, do. You’ll feel better.
Google has a different problem. It’s generally not creating content, but sorting it and delivering it. So the question of “Is this real?” and “Is this legal?” constantly come up.
In Germany, for instance, Holocaust denial is a crime (so they did learn something about the power of speech), but here any such filtering would bring howls of “suppression of free speech.” Google no doubt will double-down on AI as the solution. Its AI departments are very large and given a lot of freedom to explore unconventional ways to look at things like this. It’s appropriate that one of its premier divisions is named DeepMind.
Let’s Do Lunch
Speaking of companies that know everything about you, Amazon has started free delivery of Whole Foods groceries to Prime members in four test cities. Its target time is two hours, with one-hour service for an extra $7.99. But practically speaking, it needs to be fast, with meat and frozen food involved. I’m not sure how it’ll handle that.
What do Giant’s Peapod or Safeway’s delivery services do? Will Amazon be adding refrigerated trucks to that fleet of overly-clean white vans that you see every day — even Sunday — roaming our streets?
But you knew that once Amazon bought Whole Foods, this could not be far behind, even if Amazon’s previous attempt (AmazonFresh) died in Maryland. The reasonable, compact area of Columbia, with its existing Whole Foods, may be tempting territory for it.
Cliff Feldwich is owner of Riverside Computing and does PC troubleshooting, network setups and data retrieval for small businesses — when not wondering if your Internet-connected refrigerator soon will be ordering dinner. He can be reached at 410-880-0171 or at [email protected] Older columns are available online at http://feldwick.com.