Mark Zuckerberg’s Meta is developing a new large language model (LLM) which will compete with OpenAI’s GPT-4. The new AI model is set to launch in 2024 and will be “several times more powerful” than LLama 2. What can we hope to see from the ChatGPT rival?
Does Meta have AI?
Yes, Meta has LLaMa 2, a large language model (LLM). LLaMa stands for Large Language Model Meta AI and is the same type of AI as GPT-4 (the model that powers ChatGPT). Namely, a natural language processing (NLP) neural network. The tech giant has yet to release their own AI Chatbot, instead opting for an open-source model approach – essentially B2B in comparison to OpenAI’s B2C business model.
This is not strictly true, of course, as OpenAI has a strong footing in B2B with ChatGPT Enterprise and the ChatGPT API (featuring GPT-3.5 Turbo and GPT-4). The point here is that Meta has no B2C footing in AI chatbots at all – odd for a company of their size and specialty in 2023.
Essential AI Tools
Content Guardian – AI Content Checker – One-click, Eight Checks
Jasper AI
WordAI
Copy.ai
Writesonic
OpenAI CEO Sam Altman, creator of ChatGPT, is focused on the public-facing chatbot tool, available to the end user via web browser or the ChatGPT app. By contrast, Meta has no chat bot. Zuckerberg’s firm is instead focused on creating the AI model that other businesses use in their own third-party apps. That said, Meta does over $100B USD in annual revenue – it can afford to do both. Therefore we have no reason to expect that it wont. We can expect a ChatGPT rival from 2024 onwards.
What is Meta’s GPT-4 rival?
Meta’s GPT-4 rival is in the works, and will be a successor to LLaMa 2. The latter already being a power player, with a parameter size of 7B, 13B, or even 70B depending on which iteration you choose. Despite this, OpenAI’s ChatGPT is far in the lead, with 1.76 trillion parameters, per Mohammed Lubbad. Even GPT-4’s predecessor, GPT-3, can outperform LLaMa 2 on some tasks, with 175 billion parameters.
Now, however, OpenAI’s most advanced model will face strong competition. Meta platforms Facebook and Instagram form some of the strong data sources for human behavior – and natural human language – on earth. Where Meta owns Instagram, Facebook, and now Threads, the capabilities of OpenAI do not include access to such a fruitful data source as it’s own social media platform.
Meta CEO Mark Zuckerberg wants to “help other companies to build services that produce sophisticated text, analysis, and other output,” reports The Wall Street Journal.
Is Llama better than GPT-4?
The big AI race between major tech companies is coming to a head. LlaMa has a hard road ahead to be better than GPT-4. Top tech executives Google CEO Sundar Pichai, OpenAI’s Sam Altman, and Microsoft CEO Satya Nadella each have their own artificial intelligence system. Google owns ChatGPT rival Bard. Concurrently, Microsoft own ChatGPT competitor, Bing Chat. The creation of AI tools like LLaMa 2 rests almost entirely on the shoulders of NVIDIA, fronted by founder and CEO Jensen Huang. NVIDIA AI training chips and data centers are the hardware data infrastructure that the entire industry runs on. So when NVIDIA chips get banned, that’s bad news for everyone involved.
The world of generative AI chatbots is a natural extension of the information age in which we live. Innovation in the AI arms race is driven by a select few tech industry competitors simply because creating a new artificial intelligence system is much more expensive than most people realize. Third-party apps are relatively affordable to produce. ChatGPT rivals are abound, many of them embarrassingly built on OpenAI’s own API. However, creating the AI model itself, the neural network behind the UI, and the reinforcement training used to create it, costs many millions of dollars (at least to compete with any company mentioned on this page).
For the time being, Meta AI model LLaMa 2 is not as powerful as GPT-4. However, if rumors are to be believed, much could change in 2024.
