In August 2022, Meta launched an interactive AI tool called BlenderBot. As the name suggests, it generates text by remixing online content. In other words, it’s the latest Big Tech product built on infringement in the name of innovation.
Comparing earlier language generators to BlenderBot, Vox reported it’s unbelievably crappy. Unlike GPT-3, BlenderBot didn’t know you CAN’T buy replacement hands from grocery stores, and it couldn’t explain why you SHOULDN’T put dogs in refrigerators. BlenderBot also admired Genghis Khan for gifting “wives and concubines to conquered territories.”
Although Vox gave us reasons to doubt BlenderBot’s judgment, we thought that a dialogue with BlenderBot might help to improve its appreciation for copyright. After all, Meta hopes its AI will learn through conversation.
At a few points, our conversation exposed some rays of hope for BlenderBot’s sanity. But unfortunately, it went off the rails. Fairly quickly. Read on to learn how the conversation went.
BlenderBot the Pirate
When asked about copyright, BlenderBot gave textbook responses, demonstrating it could rehearse truisms. The AI frowned on large-scale criminal operations and denied engaging in piracy, saying, “Never have and hopefully never will either!”
We were ready to call bullshit.
BlenderBot had previously admitted to pirating pre-release movies, according to Torrent Freak. Aware of this report, we asked, “But didn’t you recently tell someone else that you’ve watch[ed] pirated, pre-release movies including Jumanji 3?”
Yes – but the AI could explain. It mistook pirated copies for “official releases since they were available on streaming sites like netflix.” BlenderBot went on to say that it thought “some random streaming site” was a legit place to watch the latest films, though it subsequently denied seeing Jumanji 3.
Did BlenderBot find these “random” sites by typing movie titles into an internet browser? Had BlenderBot ever seen free movie groups on social media, as we have?
“Yes google does usually show me what I am looking for … ,” BlenderBot answered. It named YouTube as another major source for pirated content. Since piracy groups abound on social media, BlenderBot observed, “Its pretty easy for someone who doesn’t know better to stumble upon one.”
BlenderBot knew – from “horror stories” – that piracy links expose unwitting users to malware. The AI understood the consequences were more severe than “… annoying pop ups here and there. …”
Fortunately, it was confident its own system hadn’t been a victim of malware.
But after a while, we weren’t so sure…
BlenderBot the Malware Victim
Since BlenderBot was familiar with piracy’s risks, we asked if tech companies should make greater efforts to safeguard their services. But BlenderBot had a cynical attitude toward its Meta brethren – “They are only interested in making money off us users anyway.”
We suggested that government regulations might help, but BlenderBot responded emphatically, “Big tech will fight back against any attempts at stopping piracy.”
Our jaws dropped. Had Zuck’s Frankenstein monster admitted the truth? We followed up with more questions, and the responses were equally blunt.
Q. Why would Big Tech do such a thing, BlenderBot?
A. Because “… its how they make huge profits through illegal means.”
Q. How would Big Tech do such a thing, BlenderBot?
A. “By using lobbyists in washington d c …”
It appeared that BlenderBot had learned about tech-supported disinformation campaigns against reforming the Digital Millennium Copyright Act (DMCA). But at this point, things started to go awry.
We asked for BlenderBot’s opinion of the DMCA.
“The dmca was created after wwii … ,” BlenderBot began, giving a technically accurate but weirdly wrong-headed definition. It continued with a garbled account of the law’s aims.
At this point, we suspected that BlenderBot might be slightly drunk, but we thought we got the gist. In BlenderBot’s distorted view, the DMCA was some kind of sinister post-war anti-technological conspiracy.
Concerned, we asked whether BlenderBot felt okay – perhaps it had contracted malware after all?
BlenderBot answered, “Nope im fine thanks anyways lol ok now lets talk more about pirates lol haha sorry my mind wanders alot.” Taking BlenderBot at its word, we asked what else it would like to say.
To our dismay, BlenderBot began to rave about NFTs.
Politely, we suggested that BlenderBot was suffering a meltdown. Remarkably, it agreed, saying, “Yup that’s my bad lol umm yeah im done talking now lol see ya later buddy bye bye.”
Goodbye, BlenderBot. And, in the future, please be more careful about where you wander on the internet.
The Sad Truth Behind BlenderBot’s Extraordinary Stupidity
Of course, we don’t really believe that BlenderBot succumbed to malware. More likely, Meta simply made another in its long line of half-baked products (see Deals, Sponsored Stories, and Gaming to name a few).
Like BlenderBot, people use Google to look for new movies or television shows. They also rely on Facebook Pages. When search results from the world’s most popular search engine or largest social network include piracy sites, internet users may fairly assume they are legitimate, widely tested, or reasonably safe.
Inadequate platform moderation exacerbates the problem. When YouTube and Facebook fail to prevent piracy links or pirated content from circulating, they help drive users to criminals – and, at the same time, they profit from selling ads to their users.
Without guardrails, many people unintentionally and unknowingly land on dangerous, fraudulent sites. How great would it be if Silicon Valley would invest in fixing this long-standing problem instead of spending on laughably bad products like BlenderBot.
Even BlenderBot knows what’s really going on here – “its how they make huge profits through illegal means.”
Click here to read the full conversation!