Subscribe
Share
Share
Embed
Zuckerberg's shaking off the metaverse, putting Yann LeCun in the corner, building a new line of business in cloud services, and looking for an end run around model collapse while throwing sand in the eyes of his competitors. That's what $15B and Scale AI gets you. Let's talk about it today on Minimum Viable Banter.
Creators and Guests
What is Minimum Viable Banter?
From the trenches, Technology conversation and insight, now with more... molecules.
Hello, and welcome back to Minimum Viable Banter. My name is Rod Edwards, and you can find me on LinkedIn. Search for Rod Edwards in Winnipeg or at rodedwards.ca. Today, I'm here to talk about something that I feel like is a little bit under the radar, which is Meta pumping $15,000,000,000 into sort of buying Scale AI or at least 49% of it in a deal that would also make the CEO of Scale, Alexander Wang, the head of Meta's new superintelligence AI unit. What does all this mean?
Rod Edwards:It means that Zuckerberg's shaking off the metaverse, putting Jan Lakhan in the corner, building a new line of business and cloud services, and looking for an end run around model collapse while throwing sand in the eyes of his competitors. Let's talk about it, today on Minimum Viable Banter. Thanks so much for joining me. I'd like to set the stage by starting with a question asking what exactly is Scale AI? I think there's probably a little bit of confusion or misperception in the community about what the company actually does.
Rod Edwards:The name of course makes it sound like a large language model company like a peer to Anthropic or OpenAI or something like that. But the truth of the matter is Scale is a data annotation company, which means that it's a labor company. It's a staffing organization. Scale has thousands of employees working for shell companies in low cost labor markets like The Philippines, Africa, and so forth. And their job is to tag millions of photos, classify millions of snippets of text and so forth.
Rod Edwards:This is the work of data annotation, creating the datasets that large language models are trained on. The short of it is that scale is not radically different from Amazon's Mechanical Turk product from back in the day. It's a way to provide, and I'm going to say this in sort of a buzzy way, but it's a way to provide programmatic access to human cognition. So when Scale AI describes itself as data as a service, it's because they've wrapped a layer of technology and APIs around the messy grunt work labor business of annotation. And that made Scale uniquely attractive to the AI industry just as it hits inflection point.
Rod Edwards:The stars really aligned for Alexander Wang. He was in exactly the right place and right time to pivot what started as something closer to TaskRabbit into an annotation service tailored to serve frontier language model developers and in doing so, get access to their incredible spigots of money. So two questions to address given what Scale AI is. First, what is Zuckerberg hoping to achieve? And second, why with Scale AI?
Rod Edwards:The Scale AI acquisition is only one part of what appears to be an existential push at Meta to really restore their place in the AI industry by hiring some big players. I'm sure everyone's heard about Sam Altman's response to the $100,000,000 signing offers for top AI engineers, which is absolutely bonkers. Today, there was more news about Meta making significant acqui hires from GitHub and elsewhere. Meta is spending the order of billions of dollars to capture top AI talent. That's intense.
Rod Edwards:The takeaway from that, the inevitable conclusion is that Zuckerberg is existentially scared for his core businesses. And I would also say that he's frustrated that he missed a big thing. Zuckerberg spent $65,000,000,000 and several years fiddling around in the metaverse and missed most of the generative AI gold rush. Not to say that Meta wasn't in the space, but it was in a way that was clearly not appear to the OpenAI's Anthropics and Googles of the world. In recent months, Meta's model credibility has really tapered off since LAMA four, which was widely regarded as pretty mid.
Rod Edwards:So remember for a moment who Meta's customers are. It's advertisers. Their product, Meta's product, is attention, yours and mine. And you and I are spending more time interacting with AI and less with traditional web properties and social media. At the same time, social media is getting devalued by AI slop, which creates a downward spiral that's going to impact Meta's margins, prices, and ultimately revenue and enterprise value of all traditional social media properties.
Rod Edwards:So Zuckerberg has to be feeling the heat and has correctly identified that Meta needs to be a player in the AI space to shore up its core properties. Beyond those core properties, Meta is doing another smart thing, getting into the cloud services space, hosting their own models, and making them commercially available. This feels like a good hedge against social media spending being eroded. If AI is damaging core properties, Meta can be at least the one benefiting off that cannibalization, making money off of it, and kicking off a completely new revenue stream in the fastest growing market. So to recap, why does Zuckerberg suddenly care about AI?
Rod Edwards:He's scared for his core businesses. He's aware that he's missed a window to be a peer to anthropic and OpenAI, leaving a lot of shareholder value on the table. So great. Zuckerberg is course correcting. That's great news and astute management.
Rod Edwards:And Zuckerberg is lucky because he's in a position where he's got the resources to actually throw money at course correction in a way that not many organizations would be able to muster. So the next thing is why Scale AI? First thought, maybe it's Alexander. Maybe he thinks Alexander Wang as the leader of the super intelligence unit at Meta is actually the person best equipped to win the race to super intelligence or at least artificial general intelligence. I think part of the x factor here is the mystique of Alexander Wang himself.
Rod Edwards:He's super young and super successful, and I feel like Zuckerberg must see a lot of himself in Alexander Wang. Does that make him a good bet to get to AGI or ASI? I think that's to be determined. I would argue that Alexander Wang today is a great operator. He's got a track record of smart pivots and creating an effective, fast growing product in a low margin, high headcount industry.
Rod Edwards:That's a superpower in and of itself, but he hasn't been hired to build a highly performing operations unit. He's not a frontier model developer or a model of the super intelligence Illuminati or whatever. So to be determined. This is a bet that's being placed on someone who's done some amazing things very quickly and someone that's probably very aligned with Zuckerberg's line of thought. Time will tell if that works out.
Rod Edwards:Next thought about why Scale AI is, well, it's in the name of the company. Scaling laws, I would say today, are up in the air. That is to say, what's going to power the next round of AI performance gain? Some think it's inference time. Some think it's architecture, bigger mixture of experts collections.
Rod Edwards:Others think it's the quantity or quality of data used to train a model, especially now as Internet data is increasingly locked up behind intellectual property claims and increasingly vulnerable to model collapse as AI content infiltrates everything. If the next model advances are going to be driven by data quantity or quality, then Zuckerberg is making a very smart bet, getting into bed with the foremost supplier of AI data. The other possibility, of course, is that this is throwing sand in his opponent's eyes. The cynical angle to all this is that the deal was engineered to slow down other players in their pursuit of AGI. XAI, Google, and OpenAI are all pausing work with scale.
Rod Edwards:Whatever they were doing with scale, they were doing because presumably they needed it. So, this buys time for Meta to catch up as those other tier one players are searching for new annotation partners. Also interesting to note is the KG 49% number. Meta deliberately stayed just under the margin of control, and that 49% feels finely calibrated to avoid regular regulatory oversight while scaring off other scale customers. My initial reaction to this is that I find it hard to imagine that it's not part of the equation, but the fact that it is probably feels a bit gross.
Rod Edwards:The truth of the matter is, I would say AI is firmly in sort of like Games of Thrones territory right now. Battle lines are being drawn, backs are being stabbed, lots of gross things are happening and will continue to happen. This is just sort of a more public example than what we're used to seeing. So at the end of the day, why scale? One, Zuckerberg likes the founder and probably finds them very relatable.
Rod Edwards:Two, data quality is probably really important in the next generation of models. And three, the ability to throw a wrench into the data pipeline of other tier one AI players probably has some value in and of itself. The bottom line on Scale AI is that Zuckerberg's come to terms with the fact that he's had the wrong people with the wrong priorities and has left money on the table as a result. He thinks Scale is going to plug at least some of those gaps for him and help him jump start Meta's AI factory, if not leapfrog his peers. The other thing to consider here is that Meta's market cap is close to $2,000,000,000,000.
Rod Edwards:And so a $15,000,000,000 bet on Scale AI feels like a small bet to place given the potential upside here. Those are my thoughts on Scale AI. If you'd like to connect, find me on LinkedIn or at rod edwards. Ca. And thanks so much to listening to minimum viable banter, and I'll see you next time.