Bill Jia’s Post

I started my journey with Facebook/Meta in December, 2009. After 14 years, I made a hard decision to leave Meta in Dec 2023. I joined Facebook/Meta because I believed in the company's mission: to give people the power to share and make the world more open and connected. Along these 14 years, I witnessed the company truly changed the world by turning its mission into reality. I take immense pride in being part of this journey. I still truly love Facebook/Meta and I cherish the incredible people whom I've had the privilege of working with. But after such a long time, I want to take another steep learning curve and become the driving force to accelerate another major change in the world. With this ambition, I am joining Google, supporting the Core ML/AI org. I firmly believe that AI, with a focus on safety, can bring a revolutionary change to people's daily lives. Google, being an early pioneer of the AI-first approach, is uniquely positioned to innovate in AI to enhance people's lives. I am thrilled about the opportunity to learn and be a part of this transformative journey at Google. Gemini AI ambition is too disruptive to stay just on the side and watch. Extending what I was doing in the past years (e.g. open sourcing PyTorch and creating ONNX), at Google, I want to keep contributing to the AI community in general. AI work requires a tremendous amount of intellectual research, coupled with substantial resources (tools, automation, data, etc). I think all major AI companies should take significant roles to contribute to the community via OSS. This way, we can collectively expedite the AI revolution together.

🔥🤖 The A.I. Wars are on fire! 🔥🚀 Elon Musk and Meta are stepping into the ring to claim their share of the AI boom! 💥😎 No, this won't be settled with social media posts or a cage fight (although that would be entertaining! 🥊😂). Musk and Zuckerberg's tech giants are diving headfirst into the artificial intelligence world to challenge Microsoft's OpenAI and Google! 🤯👊 Meta's got a head start, folks! 🏃♂️💨 They're gearing up to unveil an open-source version of their A.I. language model, allowing startups and developers to create their own tools and software using this cool tech! 🛠️ 🌐 According to The Financial Times, Meta's unique approach could give its competitors a run for their money! 💰 OpenAI and Google, on the other hand, have been keeping things hush-hush with their proprietary systems. 🤫🔒 But Meta is like, "Hey, let's all collaborate and make this AI party even better!" 🎉🙌 They're all about industrywide teamwork to calm down those pesky tech fears! 🙏😱 And guess what? Meta has been playing with A.I. for ages! 😎 They've invested billions into it, and way before the metaverse, they cooked up LLaMA, a powerful chatbot fuelled by A.I. technology! 🤖 💬 Meta's chief A.I. scientist, Yann LeCun, is considered one of the "godfathers" of A.I., but he's got a fresh take! He's like, "Existential threats? Nah, they're ridiculous!" 😅🤷♂️ So buckle up, folks! The A.I. arena just got even hotter with Elon Musk and Meta throwing down the gauntlet! ⏳ 🔥 Who's ready to witness the battle of the tech titans? Let the games begin! 🏆🎮 #ai #techwars #elonmusk #meta #openai #google #artificialintelligence #artificialintelligencetechnology

Mark Zuckerberg explained how Meta will "crush Google and Microsoft at AI"—and Meta warned it could cost more than $30 billion a year With an amalgamation of vast data reservoirs, gleaned from users across its apps, and a pioneering approach to disseminating ML model weights, I do believe that some of the important foundations are in place and therefore the odds could indeed be favourable. But its hard to predict what the "competition" will do next month, let alone over the next 2 or 3 years which is the timeframe Meta is talking about when explaining its approach to investors. The decision to provide developers with access to model weights, albeit not truly "open-source" by definition, has captured a huge chunk of innovative and influential grassroots users. When ChatGPT came out, people were making YouTube videos about how to use it in different ways. If you check now, the vast majority of these kinds of videos are based on the Llama model and its derivatives. And that's saying something. It is a big change in a short time. However, three things remain constant, to paraphrase some thought leaders, so over-confidence is unwise: 1) Predictions are uncertain, especially when they pertain to the future 2) In AI and Big Tech, anything can happen, and it usually will 3) What got you here, won't always get you there #LLM #opensource #GenAI #AGI #walledgarden #FANG Source - https://lnkd.in/exfgugMu

Google was able to build #gemini with strong #multimodal capabilities largely because of its search data, #youtube and #gmail. what do you think Meta's capability of building an equally good or even better #LLM with data from platforms like Facebook, Instagram and WhatsApp that they own? (If you want to have a bit more fun, ask the question in both #Bard and #ChatGPT and see if you agree with the responses 😃) Are we giving too little credit to what Meta can do in the #GenAI race while all the spotlight is on Microsoft and Alphabet Inc.? A hint - Meta is the largest buyers of NVIDIA #GPUs, as many as Microsoft and 3x Google. Yes Meta has given the world an #opensource model. But if we assume Meta isn't far behind in hiring and keeping top AI talents, my bet is what they've been building internally must be amazing. If you have any insights, please share.

As a key player in the technological arena, Meta is unaccustomed to trailing behind. Yet, in the artificial intelligence (AI) sphere, they find themselves wrestling with an unfamiliar reality—pursuing the pack. In spite of having pioneered AI, Meta is losing ground in this critical technology, even as competitors charge ahead. This dichotomy is particularly pronounced as they strive to channel resources into market-ready AI products, including chatbots, moving from a foundation of academic exploration to practical, commercial application. Meta's trials are magnified by a recent exodus of AI talent and a series of layoffs, coupled with increasing executive oversight of AI strategy. The company's relevance, growth, and competitive edge are all under threat, as it struggles to match the pace of innovative AI startups and industry leaders like OpenAI and Google. It is a pivotal moment for Meta—can it commercialize its AI efforts, enhance user engagement, and appeal to a younger demographic in the face of these challenges? In a telling example of the challenges Meta faces, November 2022 saw the company release Galactica, a large language model with a focus on science. However, the system was abruptly taken down by Meta within mere days of its launch, following a flurry of criticism from the scientific community over its inaccurate and biased responses. The troubles of this AI endeavor underscore the tech giant's struggles and put the industry's fast pace into stark relief, especially considering what happened next. As if to punctuate Meta's difficulties, just two weeks later, OpenAI unveiled ChatGPT. https://zurl.co/XGFR

As a key player in the technological arena, Meta is unaccustomed to trailing behind. Yet, in the artificial intelligence (AI) sphere, they find themselves wrestling with an unfamiliar reality—pursuing the pack. In spite of having pioneered AI, Meta is losing ground in this critical technology, even as competitors charge ahead. This dichotomy is particularly pronounced as they strive to channel resources into market-ready AI products, including chatbots, moving from a foundation of academic exploration to practical, commercial application. Meta's trials are magnified by a recent exodus of AI talent and a series of layoffs, coupled with increasing executive oversight of AI strategy. The company's relevance, growth, and competitive edge are all under threat, as it struggles to match the pace of innovative AI startups and industry leaders like OpenAI and Google. It is a pivotal moment for Meta—can it commercialize its AI efforts, enhance user engagement, and appeal to a younger demographic in the face of these challenges? In a telling example of the challenges Meta faces, November 2022 saw the company release Galactica, a large language model with a focus on science. However, the system was abruptly taken down by Meta within mere days of its launch, following a flurry of criticism from the scientific community over its inaccurate and biased responses. The troubles of this AI endeavor underscore the tech giant's struggles and put the industry's fast pace into stark relief, especially considering what happened next. As if to punctuate Meta's difficulties, just two weeks later, OpenAI unveiled ChatGPT. https://zurl.co/XGFR

🚀Did Meta just do what OpenAI was originally set out to do?!?! And how does Watson play into this? 🚀 I often ask people if they know the key difference between Watson (circa 2011) and OpenAI (circa 2022). The simplest answer is Watson was B2B, and OpenAI went DTC with the goal of democratizing AI capabilities to the masses. However, when OpenAI decided to: 🧘♀️ Pivot from a non-profit to for profit company 🧘♀️ Charge for this ‘democratized’ AI capability So now they’ve become like ‘Watson’ but with a low-price, low-entry offering to consumers, which has lead to universal concern that their well-intentioned start had veered off the deep end resulting in removal of the CEO (ouch!) and Elon not being happy (resulting in legal affairs). Oddly enough, there’s a strange correlation with the path OpenAI is on and with how Meta started. Meta was for profit, first to market, and got a heap of backlash for it because they started the social movement (in comes OpenAI and what they did with ChatGPT). However, Meta is desperately trying to rebrand themselves by: 🧘♀️ Changing their name to ‘Meta’, no longer the big bad empire that is Facebook 🧘♀️ Releasing Llama 3, the strongest performing open-sourced LLM, showcasing their appeal to the open-sourced community. Is it possible that Meta is changing gears and doing what OpenAI was originally funded to do???

Breaking News: The world of #martech is set to take a giant leap forward following a surprising announcement from Meta's CEO, Mark Zuckerberg. Today, Zuckerberg revealed that Meta is making Llama 2, the next generation of their open source large language model, available for free for both research and commercial use. Partnership with Microsoft Remarkably, Meta is driving this innovation by partnering with Microsoft. Users can access the new Llama 2 models via Microsoft's Azure cloud service. The synergy between two tech titans also yields a locally optimised version for Windows users. Zuckerberg stating he did so driven by a belief in the power of open-source technology. He underlines that it drives innovation by allowing a multitude of developers to leverage new technologies. This development is potentially groundbreaking for businesses. Meta grants commercial access to Llama 2, providing startups and other companies a powerful, cost-free alternative to models sold by AI providers like OpenAI and Google. While the AI software industry has been dominated by proprietary models, Meta's decision to make Llama 2 widely available and free could potentially disrupt the market as we know it. This move not only affords the opportunity for more developers into the AI space but also potentially strategically positions Meta as a benefactor for future advancements within the industry. Meta does have a track record of open sourcing their AI and infrastructure work. Notable instances include PyTorch, one of the leading Machine Learning frameworks and models like Segment Anything, ImageBind, and Dino. With Llama 2. Llama 2 boasts significant improvements in safety, quality, and performance. It was trained on 40% more data than its predecessor and fine-tuned using over 1 million human annotations, supervised fine-tuning, and reinforcement learning with human feedback. As we navigate this new era of AI in martech, we at MartechFest are beyond thrilled. The implications of this partnership between Meta and Microsoft, along with the release of a powerful, open-source AI language model like Llama 2, will present incredible opportunities for innovation in our field. We’re excited about the creativity, fresh perspectives, and cutting-edge solutions that will undoubtedly emerge from this pivotal moment. As Mark Zuckerberg stated: “Open source drives innovation because it enables many more developers to build with new technology.". And indeed, this move could unlock massive progress if even more of the tech and AI ecosystem embraces the openness that Meta states they envisage. Stay tuned for more updates, and in the meantime, keep innovating! #AI #meta #techstack

Meta will release Llama 3 in July; such are the rumors. How should Meta avoid the kind of public relations disasters that Google is experiencing? The easy answer is Optional Guardrails. Llama 3 is the highly anticipated follow up to Llama 2, the AI model that made open source a viable alternative to BigAI. Meta took a safe approach to the Llama 2 release in order to ward off the kinds of disastrous backlash Google is experiencing now. I've thought that there should be very limited guard railing in AI, much like there was with the early Internet. On the early Internet you could find absolutely anything, both wonderful and wicked. Over time, however, humankind cleaned up the Internet and relegated much of the wickedness to places like the dark web. We should have done the same with AI because speech is an historically difficult (near impossible) area to regulate. With AI it'll be even more difficult. We need to let some things play out before trying to anticipate absolutely everything that could go wrong. AI isn't a Starship. We won't all die if something unexpected happens. But no matter what Meta does to mitigate it, the upcoming Llama 3 release is going to be closely watched for opportunities to embarrass Meta. If they make Llama 3 super safe, it'll do and say stupid things and embarrass Meta. If they completely remove the guard rails, they'll be accused of trying to kill humanity. To navigate the dichotomy, Meta should release two Llama 3's: one with all the guards and protections any corporate HR department head could dream of. And another with no guardrails at all; only a warning: "This AI might say anything at all. Beware." And then enjoy the ride. That way no one is happy, and the status quo is preserved. So, how should Meta handle Llama 3? Optional Guardrails is the answer.

Should the big kids on the playground set the rules? Do we really trust corporations who have fiduciary duties to regulate their own interests? The recent alliance forged between Meta and Microsoft, along with their active participation in the "Responsible Practices for Synthetic Media: A Framework for Collective Action" initiative by Partnership on AI, prompts inquiries into their underlying motives. While they present this collaboration as a drive to endorse ethical AI-generated content, a thorough examination becomes imperative to ascertain whether these tech giants genuinely champion the welfare of the general populace or if their involvement serves strategic posturing purposes.   Given their historical controversies and ethical dilemmas, it becomes crucial to closely evaluate Meta and Microsoft's underlying intentions. Can we place confidence in their dedication to safeguarding users and combatting misinformation, or does this engagement primarily serve as a veneer to enhance their public perception?   Although their participation could lend credibility, it simultaneously raises apprehensions about the concentration of power. Will this partnership foster inclusive collaboration and embrace a wide range of viewpoints, or will it instead serve as a tool for these corporations to exert undue influence?   To establish authentic accountability, the Partnership on AI must uphold transparency and actively involve impartial experts and civil society organizations in the decision-making process. Rigorous scrutiny of implementation strategies and practical recommendations is paramount.   While Meta and Microsoft's involvement signals a positive stride, it remains crucial to remain vigilant and demand transparency. Only by consistently prioritizing responsible practices and confronting ethical complexities can we genuinely believe in their commitment to safeguarding the collective welfare. #ethicalAI #womenentrepreneurs #selfregulation #microsoft #meta