4 ms·
You know Meta has the best tech in terms of AI/ML right? They just need a new leadership focus to steer the company in the right direction. They literally have
by lvl102 4y ago
You know Meta has the best tech in terms of AI/ML right? They just need a new leadership focus to steer the company in the right direction. They literally have all the tools to lead in that area.
- rcme 4y agoDo they? The LLM they released was much worse than GPT-3.
- lvl102 4y agoAll those models are only as good as the data they’re trained on. Google and Meta both have the most relevant proprietary data that cannot be rivaled. Not even close.
- vernon99 4y agoThis depends on what are you training the model for. Arguably wikipedia is a much better dataset for all things related to real world and history than Fb’s feed or activities or a bunch of websites optimized for SEO. This is easily illustrated by ChatGPT not being Google or Meta and doing a pretty great job with available sources. So no, this doesn’t matter much, imo.
- roland35 4y agoAs an encyclopedia sure wikipedia is great but the type of data Google and Meta has is totally different. Basically a real time feed of who is talking to who about what, and what people are searching for. Totally different applications
- lvl102 4y agoThat’s incredibly naive. ChatGPT is not the state of the art. Not even close.
- kibwen 4y ago> All those models are only as good as the data they’re trained on. Having seen the average level of discourse on Facebook, I have no reason to believe that it contains any proprietary data worth training on.
- coffeebeqn 4y agoI guess you could train the ultimate toxic shitposter and then do the opposite of what it says
- celestialcheese 4y agoIt's largely worse from the outside because meta has no goodwill to make mistakes and iterate in public. GPT-2/3 were just as bad at hallucination, if not worse. That doesn't mean that meta isn't iterating on their LLM in private using the 3b users and armies of labelers to fine-tune and improve their model. From what I understand, they have one of the strongest translation models at scale, and their recommendation models for advertising are likely only rivaled by Google or Bytedance
- gen220 4y agoHow does this concretely translate into a marginal $30bn/year in revenue? Inverting the question, who would collectively pay $30bn/year to buy "AI/ML from Meta", assuming you're talking about a new product offering.
- deltree7 4y agoThere are only two platforms where AI/ML will be delivered. Cloud for the Enterprise and Mixed Reality for People. Mixed Reality is the final platform that kills all other platforms and has a potential greater than $40 Trillion
- germinalphrase 4y ago“ There are only two platforms where AI/ML will be delivered” Why would you believe these are the only platforms?
- deltree7 4y agoIn a broad investment sense. Sure there will be other niches. But just like Mobile phones and Servers were the twin tech platforms till 2020 that delivers all of digital value. That paradigm will shift to MR and Cloud. If we are all going to live in a Metaverse, you need a MR device to deliver that experience to the user and Cloud to centralize/distribute those experiences. Besides food (and physical healthcare) and a climate-controlled 20 x 20 space, you don't need the real world (atoms). All your experiences will be delivered through bits.