Want to give full credit to O’Reilly Radar and NYTimes.com for the articles they published that got me thinking about threats to the veracity and usefulness of the Internet these days.
SIFT and DeepCheapFakes
Tuesday, May 11, 2021
I know Mike Caulfield was spotlighted in a NYTimes article addressing ways to defend oneself against the tricks and fakery of Conspiracy Theory promoters
SIFT (The Four Moves) has been a clarion call to help arm students against the vagaries and intangible hidden motives of the Internet-at-Large. https://hapgood.us/2019/06/19/sift-the-four-moves/. But can one laterally read “at scale”? I think the argument against that would be, whaddya mean? These are “individual” articles and claims made on individual websites. But even Caulfield understands and has said in past that it is so hard to defend against this because it’s so easy to create the disinformation in the first place. This isn’t just about the attention one can devote to performing the “The Four Moves”, this more like Nuclear Proliferation. We’re not subject to a finite set of people who are creating content in bad faith. No. We are dealing with people who have access to tools, akin to nuclear warheads and the industrial complex (meaning Nuclear Power Industry), that can provide large amounts of precious materials “at scale” to make warheads. This is where the U.S. and U.S.S.R. found themselves in the early 1980s, with 20,000 warheads apiece. If you have the nuclear power plants generating power, it’s easy to process the spent fuel rods into Tritium gas and Plutonium 239, it’s a mere extraction process. Before you know it, you have more raw material than you could ever dream of to create a stockpile of weapons to use. Thankfully we don’t use them, but the danger is real. Even today.
Which brings me to the statement of disinformation, bad-faith actions “at scale” when it comes to navigating sources of so-called information on the Internet. In the NYTimes article the four moves are SIFT
2. Investigate the source.
3. Find better coverage.
4. Trace claims, quotes and media to the original context.
Luckily in the process, there’s always some good faith actors, source of record with copies end-notes with further references (those are harder to fake). But wither Wikipedia? And news organizations of recorder (in an example provided to the NYTimes report, news service Agence France-Presse fact-check site is used). In the article does admit later, that “SIFT is not an antidote to misinformation”. But still there’s a further lurking danger out there. And it isnt’ the depletion of our attention, and what to direct it at. It’s the casual, CheapDeepFake as described in O’Reilly Radar website (of Tim O’Reilly fame). https://www.oreilly.com/radar/deepcheapfakes/ by Mike Loukides
So while Mike Caulfield is arguing it’s our attention span that’s being gamed by these sources of bad faith, disinformation on the Internet, Mike Loukides is pointing out the “at scale” threat, the nuclear proliferation model of disinformation. Meaning, what if one can casually create seemingly legitimate responses, valid comments to public calls for citizen input? The prime example is the recent discovery that FCC requests for public comments on elimination of Net Neutrality was “gamed” most likely using at technology, tool to create disinformation “at scale”. No amount of attention could sift through the use of legitimate people’s names unlawfully for fake public comments on the FCC website. Furthermore, now there are trained AI engines who specialize in human readable text (GPT-3) https://en.wikipedia.org/wiki/GPT-3https://en.wikipedia.org/wiki/GPT-3. Which can be “wielded” for a low rate per hour from Cloud Services providers like Microsoft’s Azure. This is what Mike Loukides is referencing when he is worried about DeepCheapFakes. The quality of output of a GPT-3 instance is “so high that it is difficult to distinguish from that written by a human”. So while you may do a SIFT Four Move search on Google, read laterally, how will that “scale”, how will your attention smart bomb deal with the overwhelming response of GPT-3 generating effortlessly millions of bits of content, at-scale, DeepFake and in bad faith? One can only hope that like Nuclear Weapons everyone will be afraid to use them, though the consequences may be less dire (end of civilization, end of the planet vs. the end of Civil Discourse). But the problem is still the same. In past books, TV, Film, were all inaccessible to the majority and tended to favor through things like editors, and peer reviewers. So there was a built-in attention economy favoring good faith actors in those media. But not-so the Internet. Nay, we are faced with new inventions, new weapons of casual destruction, fast, cheap and outta control for whatever ends the bad-faith actor may pursue. The only way to not fall victim, sadly, just as was the case in the movie War Games, is to “never play the game”. Meaning, one should consider the inherent value of the Internet. If this is what is possible, what it has come to. My attention will be re-directed, and analog. But for now I put my money where the source is, I fund things like Wikipedia as it’s a bulwark against the causal DeepCheapFake (for now at least, until GPT-3 compromises that set of peer reviewers and editors). Fingers-crossed.