April 19, 2023
Fun fact: Vancouver was once called Gastown, and Granville after that. Anyway, today's Login is 1,208 words, a 5-minute read.
1 big thing: The global elite is excited and terrified by AI
When the TED audience was asked Tuesday whether they were excited by artificial intelligence, most people raised their hands. When asked whether they were scared about AI, most people also raised their hands.
Why it matters: This enthusiastic ambivalence reflects society's broader split over a rapidly advancing technology that both tantalizes and terrifies.
Driving the news: AI has dominated the early part of this year's conference in Vancouver, with Tuesday morning's talks highlighting both the exciting promise and potentially apocalyptic future that the technology could portend.
The positive case
- OpenAI co-founder Greg Brockman showed a series of demos of what's coming in the near future. In one, ChatGPT suggested a post-TED meal, used a DALL-E plugin to visualize the meal, created a shopping list on Instacart and then tweeted out that list.
- In another example, ChatGPT dissected a spreadsheet to suggest several ways to display data within the program, and then graphed them.
- Sal Khan showed off Khan Academy's work to turn GPT into a useful tutor and teacher's assistant, a subject we covered in Login recently.
Sounding the alarm
- Eliezer Yudkowsky, who argues modern AI development needs to be shut down, highlighted the existential threat posed by the imminent arrival of machines built by humans that can have superhuman intelligence and act in ways humans don't fully understand.
- Even if we don't know exactly how such systems might cause human extinction, he says the risk is high.
- "I suspect we could figure out with unlimited time and unlimited retries," he said, but insists that's not the situation. "We do not get to learn from our mistakes and try again."
AI is a mixed bag
- University of Washington professor Yejin Choi argued that AI systems need to be re-architected and taught both common sense and human values, both of which she said are severely lacking in even the latest large language models.
- Choi believes that building ever bigger models alone won't solve these fundamental limitations. "You don't reach to the moon by making the tallest building in the world one inch taller at a time," she said.
- Scale AI CEO Alexandr Wang made an impassioned case that U.S. and its allies must harness AI for military use faster than its adversaries. AI, he said, is already changing the nature of warfare, from weaponizing drones to enabling disinformation and cyberattacks on infrastructure.
- "The AI war will define the future of our world," Wang said. "We cannot sit on the sidelines and watch the rise of an authoritarian regime. We must fight for the world we want to live in."
Between the lines: Less addressed during Monday's talks were the more subtle challenges posed by AI, including how AI will impact jobs or the potential for it to disproportionately advantage the already powerful and wealthy while doing harm to those already marginalized.
Be smart: AI's benefits could extend well beyond the automation of mundane tasks to help address challenges that humans have struggled with, from climate change to curing disease.
- But this is powerful technology and it will inevitably be used for both good and bad. Identifying the good that AI can do won't, by itself, make any of its harms disappear.
The bottom line: Smart people in the field say they share both the optimism and fears being voiced. "We hear from people who are excited. We hear from people who are concerned," Brockman said. "Honestly that's how we feel."
2. Platforms pull "fake Drake" song made with AI
A viral AI song that replicates Drake and The Weeknd's vocals has been pulled from multiple streaming platforms following a complaint from label Universal Music Group (UMG).
Driving the news: "Heart on My Sleeve" was removed from Spotify and Apple Music on Monday. By Tuesday, it was removed from YouTube, Amazon, SoundCloud, Tidal, Deezer and TikTok — where it was streamed 15 million times after being originally uploaded on the platform by a user called Ghostwriter977. Some versions were still available.
- Ghostwriter977 said in a post that they wrote and produced the rap song using artificial intelligence.
By the numbers: The song that was played 600,000 times on Spotify and attracted 275,000 views on YouTube was widely shared on social media, with one clip posted to Twitter garnering nearly 20 million views before the media was disabled.
What they're saying: UMG said in a media statement: "The training of generative AI using our artists' music" represented "both a breach of our agreements and a violation of copyright law."
- The music publisher added that platforms had a "legal and ethical responsibility to prevent the use of their services in ways that harm artists."
Reality check: While there are intellectual property issues, it's not really clear whether the label or Drake and The Weeknd have a claim that the song itself violates copyright law, given that it isn't something the artists ever wrote or sang.
- A second issue is whether the engine that created it violated copyright law, assuming it was trained on Drake's music.
- A number of media companies across multiple content types have said that they believe that AI engines are violating the law when they use copyrighted material as training data.
Our thought bubble: This is just the beginning of what's likely to be a long and complex conflict between excitement over deploying AI in pop culture and efforts to protect copyright.
Go deeper: Generative AI is a legal minefield
3. Netflix stops mailing DVDs
Netflix on Tuesday said it's ending its 25-year-old mail-in DVD business and delaying plans to crack down on password sharing more broadly until later this year, Axios' Sara Fischer and Hope King report.
Why it matters: The streamer has been looking closely at costs as its growth slows amid greater competition.
- "Converting moochers into paying subscribers," as J. Clara Chan at The Hollywood Reporter put it, is one way to drive revenue growth.
- Buying new DVDs has been a big recurring cost for Netflix, while making up just 0.5% of its total revenue.
Details: Netflix said that paid account-sharing this year in Canada, New Zealand, Spain and Portugal has initially led to cancellations, but that over time, people ultimately activated their own accounts or added new members.
- "We found enough improvement opportunities in these areas to shift a broad launch to Q2 to implement those changes," the company said in a statement.
- As for its DVD business, Netflix says it will conclude on Sept. 29, with final shipments.
4. Take note
- TED2023 continues in Vancouver.
- Meta is expected to lay off more employees Wednesday, with Vox reporting some 4,000 could lose their jobs. These would be part of the 10,000 cuts announced in March, which followed 11,000 layoffs in November.
- IBM is slated to report earnings after the markets close.
- Kiersten Todt, chief of staff at the Cybersecurity and Infrastructure Security Agency, is returning to the private sector and will be replaced by Kathryn Coulter Mitchell, currently a deputy undersecretary at the Department of Homeland Security.
- Seventeen House members and 585 congressional aides were affected by a data breach at the D.C. health insurance marketplace last month, the marketplace's director will tell Congress on Wednesday. (Axios)
- Twitter has quietly removed the section of its hateful conduct policy that prohibited deliberate misgendering and deadnaming of transgender and nonbinary people. (The Verge)
5. After you Login
Meet Pearl, the world's shortest dog.
Thanks to Scott Rosenberg and Peter Allen Clark for editing and Bryan McBournie for copy editing this newsletter.