Full Transcript
This transcript was prepared by a transcription service. This version may not be in its final form and may be updated.
Alex Ossola: Welcome to Tech News Briefing. It's Friday, April 19th. I'm Alex Ossola for the Wall Street Journal. Coming up on today's show, Meta has come out with its latest artificial intelligence model. We'll hear about what's new and how it fits into the company's apps from WSJ social media business reporter Salvador Rodriguez. And then Grayscale's Bitcoin fund is the largest in the world, but now investors are fleeing in favor of its lower-cost competitors like BlackRock and Fidelity. WSJ reporter Vicky Ge Huang tells us how Grayscale is fighting to compete with Wall Street giants. But First, Meta has released its latest publicly available AI model called Llama 3. To start it's releasing two versions, an 8 billion parameter model, and a 70 billion parameter model. Parameters are variables in an AI system that measure a model's size and capabilities. So how does Meta's new model stack up to competitors like ChatGPT and Google Gemini? WSJ social media business reporter Salvador Rodriguez is here to tell us. Sal, where might the users of Meta's products encounter this new AI model?
Salvador Rodriguez: Yeah, so before it was a little bit hidden. Now what they're going to do is that they're putting Meta AI front and center within the search function of the various apps. And so now whenever you go to search anything, you should be encountering Meta AI at that point.
Alex Ossola: Tell me a little bit about how Meta created this model. Was it different than models in the past?
Salvador Rodriguez: From what we saw in the materials that they sent over, what stuck out is that they said that for Llama 3, they used synthetic data to train the model. So what does that mean? Essentially, these models guzzled through so much data that they've actually run out of things to read. Humans have just not created enough text content for them to be able to get the training that they need. And so essentially now, one way to get ahead in this AI race is to figure out how do you solve for the lack of data problem. And what Meta is saying by telling us that they're using synthetic data is that essentially AI models are creating data, which is then being used to train the new AI models.
Alex Ossola: Llama 3, like its predecessors, is open-sourced. What exactly does that mean for an AI model?
Salvador Rodriguez: That means that if other folks on the internet want to use Llama 3 to build their own creations, they should be able to do that. Whereas if you were using the models from other companies, you have to pay. So that's one difference for Llama 3 is that Meta does put this out there for free.
Alex Ossola: What does Meta's new model mean for companies like OpenAI and Google that are making competing models?
Salvador Rodriguez: It raises the bar for them. Because if Meta is going to have a version of these large language models that's out there that seems to get the job done for free, then the bar for charging goes up higher. And Meta isn't the only company that's put out an open source model, but they do seem to be the largest player that appears to be dedicated to the open source tactic. That said, they did mention to us that when the 400 billion parameter model comes out, that perhaps they would consider a version where they do charge for use of the larger versions of Llama 3, just because these things are very expensive to run. So perhaps that'll happen, but at least at the 8 billion and 70 billion parameter level, it is an available option for free.
Alex Ossola: Why does Meta put out its models for free?
Salvador Rodriguez: They're the ads company, right? That's how they make their bread and butter. They've really got that market cornered. But what they've always said is that because of those ads, they can offer their products for free. And that is true with Facebook and Instagram and everything. So essentially, when it comes to LLMs, their approach is in line with their history, and they want to take this new breakthrough technology in the field of AI and democratize it, meaning make it available to their users for free. So are these tools going to be best in class? I don't think so, but the angle that they're going for is best in class amongst the free versions.
Alex Ossola: That was our social media business reporter Sal Rodriguez. For more of your daily tech headlines, check out TNB Tech Minute. That's right here in the Tech News Briefing feed. Coming up, why are investors pulling money out of the world's largest Bitcoin fund? That's after the break.For years, the Grayscale Bitcoin Trust was one of the few ways to invest in Bitcoin without buying the cryptocurrency itself. But after the Securities and Exchange Commission approved Bitcoin ETFs in January, more than a dozen competing funds have popped up, including ones offered by Wall Street giants like BlackRock and Fidelity Investments. Now, the Grayscale Bitcoin Trust, which is still the world's largest Bitcoin fund, is bleeding. Its assets have shrunk by more than half from their peak in November 2021. WSJ reporter Vicky Ge Huang joins us now to tell us why. Vicky, what is Grayscale Bitcoin Trust? How is it different from Grayscale Investments?
Vicky Ge Huang: So the Grayscale Bitcoin Trust is a product through which you can invest in Bitcoin in your brokerage accounts, and it's a product of Grayscale Investments, the crypto asset manager. And the fund was for many years, the first and the only product approved by the Securities and Exchange Commission for investors to invest directly in Bitcoin instead of through futures contracts or directly buying Bitcoin on a crypto exchange. But it occupied an interesting role because it wasn't an exchange-traded fund before. It is an exchange-traded fund now. So it tracks Bitcoin's price, which means that the market price of the Grayscale Bitcoin Trust right now is very much close to or almost identical to Bitcoin's price. However, in the past, it traded sometimes at a premium to the price of Bitcoin or sometimes at a discount to the price of Bitcoin.
Alex Ossola: You write that the trust assets have shrunk by half from their peak in November 2021. What's going on there?
Vicky Ge Huang: For the longest time, a lot of investors put a lot of money into the trust seeking an easy and frictionless way to invest in Bitcoin. So the assets in the fund grew, and because the trust traded at a premium to the price of Bitcoin for a long time, that also attracted a lot of investors to put money in the fund betting that perhaps this premium could continue. However, Grayscale converted the Bitcoin trust into an exchange-traded fund, meaning that investors can take their investments out of the fund without having to suffer a discount on the value of their shares. So this resulted in a lot of investors taking money out of the fund because, since early 2021, the Grayscale Bitcoin Trust had been trading at an increasingly steeper discount to the value of the Bitcoin it held. So those investors obviously wanted out immediately.The other aspect of this is that a lot of hedge funds and investors, they also put a lot of money into the Grayscale Bitcoin Trust when it was trading at a discount to the price of Bitcoin because they saw it as a trade where they could profit when the discount narrows. And then the last aspect of this is that a lot of crypto companies that collapsed in 2022, they held a lot of Grayscale Bitcoin Trust shares, so because they are bankrupt, and they needed to liquidate their holdings to pay back creditors. So a lot of them, like FTX, they sold all of their GBTC shares.
Alex Ossola: So for investors who are pulling their money out, where are they taking that money instead?
Vicky Ge Huang: In mid-January, when the Grayscale Bitcoin Trust converted into an exchange-traded fund, nine other competing funds launched at the same time. These are brand new Bitcoin exchange-traded funds launched by the likes of Wall Street asset managers such as BlackRock, Fidelity Investments, and many other players. And it seems like investors, a lot of them have been putting money into funds launched by these Wall Street asset managers because a lot of them are charging very low, cheap fees and sometimes close to nothing as a promotional fee waiver.
Alex Ossola: Asset managers like BlackRock and Fidelity seem to be charging lower fees than Grayscale Bitcoin Trust. How has Grayscale responded?
Vicky Ge Huang: They filed to launch a smaller fund called the Grayscale Bitcoin Mini Trust. That's tied to the Grayscale Bitcoin Trust where it could offer investors lower fees. And if this fund were to be approved by the SEC, each current GBTC shareholder's holdings of the trust, a certain percentage that's not disclosed, would be transferred to this smaller mini trust.
Alex Ossola: What has Grayscale said about this? Are they optimistic about their future prospects?
Vicky Ge Huang: From chats with the CEO and industry watchers, Grayscale is still very optimistic about their outlook, and a lot of it has to do with this crypto bull market that we are seeing right now. Bitcoin's price has surged a lot from just earlier this year before the Bitcoin ETFs launched, and that has presented a rising tide lifts all boats effect on the crypto industry.
Alex Ossola: That was our reporter Vicky Ge Huang. And that's it for Tech News Briefing. Today's show was produced by Julie Chang. I'm your host, Alex Ossola. Jessica Fenton and Michael LaValle wrote our theme music. Our supervising producer is Katherine Milsop. Our development producer is Aisha Al-Muslim. Scott Saloway and Chris Zinsli are the deputy editors and Philana Patterson is the Wall Street Journal's head of News Audio. We'll be back this afternoon with TNB Tech Minute. Thanks for listening.
