Big tech wants to make AI cost nothing
Meta has open-sourced its Llama 3.1 language model for organizations with fewer than 700 million users, aiming to enhance its public image and increase product demand amid rising AI infrastructure costs.
Read original article
Meta has recently open-sourced its powerful Llama 3.1 large language model (LLM), allowing organizations with fewer than 700 million monthly active users to utilize it freely. This move raises questions about Meta's motivations, which may include improving its public image amid privacy concerns and following a classic Silicon Valley strategy of commoditizing complements. By making LLMs more accessible, Meta aims to increase demand for its products, similar to past strategies employed by Microsoft and Google. The release comes at a time when the costs of AI infrastructure are soaring, with estimates suggesting that recouping recent AI spending could exceed $600 billion. Despite the technical challenges of training such models, Meta plans to significantly scale its capabilities, potentially producing hundreds of advanced models annually.
Other tech giants like NVIDIA, Microsoft, and Google are also releasing LLMs, which could threaten the business models of smaller AI startups like OpenAI and Anthropic. The natural complement to LLMs is server capacity, which cloud providers stand to benefit from as they can rent out resources for running these models. Meta's strategy may also focus on user-generated content, enhancing engagement on its platforms. The ongoing commoditization of LLMs could lead to a reckoning for smaller companies, while the massive infrastructure build-out in AI may pave the way for breakthroughs in various fields, including robotics and drug development.
Related
Meta AI develops compact language model for mobile devices
Meta AI introduces MobileLLM, a compact language model challenging the need for large AI models. Optimized with under 1 billion parameters, it outperforms larger models by 2.7% to 4.3% on tasks. MobileLLM's innovations include model depth prioritization, embedding sharing, grouped-query attention, and weight-sharing techniques. The 350 million parameter version matches larger models' accuracy on specific tasks, hinting at compact models' potential for efficiency. While not publicly available, Meta has open-sourced the pre-training code, promoting research towards sustainable AI models for personal devices.
OpenAI slashes the cost of using its AI with a "mini" model
OpenAI launches GPT-4o mini, a cheaper model enhancing AI accessibility. Meta to release Llama 3. Market sees a mix of small and large models for cost-effective AI solutions.
Meta won't release its multimodal Llama AI model in the EU
Meta will not release its Llama AI model in the EU due to regulatory concerns, impacting European companies. Apple also considers excluding the EU from its AI rollout. This decision poses challenges for companies worldwide.
Llama 3.1: Our most capable models to date
Meta has launched Llama 3.1 405B, an advanced open-source AI model supporting diverse languages and extended context length. It introduces new features like Llama Guard 3 and aims to enhance AI applications with improved models and partnerships.
Meta releases an open-weights GPT-4-level AI model, Llama 3.1 405B
Meta has launched Llama 3.1 405B, a free AI language model with 405 billion parameters, challenging closed AI models. Users can download it for personal use, promoting open-source AI principles. Mark Zuckerberg endorses this move.
14 commentsBig tech wants to make AI cost nothing to end-users maybe, but Google and Microsoft want the cost of hosting AI to make your eyes bleed so you don't compete or trim any profits off their cloud services. As the article points out Facebook does not offer cloud services so its interests in this case align with mom and pop shops that don't want to be dependent on big tech for AI.
But Mistral was way more useful to mom and pop shops when they were trying to eke out performance from self-hostable small models. Microsoft took them out of that game. These enormous models may help out boutique data center companies to compete with big tech's cloud offerings but it's beyond a small dev shop who wants to run a co-pilot on a couple of servers.
Microsoft and Google don't want you to learn that a 7B model can come close to a model 50x-100x its size. We don't know that's even possible, you say? That's right we don't know, but they don't even want you to try and find out if it's possible or not. Such is the threat to their cloud offerings.
If they did Microsoft would have made a much bigger deal of things like their Orca-math model and would have left Mistral well alone.
Deepseek v2 lite is a damn good model that runs on old hardware already (but its slow).
In 2-3 years we will likely have hardware that runs 70b parameter models with enough speed that you will run it locally.
Only when you have difficult questions will you actually pay.
For example I already use https://cluttr.ai to index my screen shots and it costs me $0.
(I made this tool tho)
Generalizing somewhat but focusing on a single company:
https://www.statista.com/statistics/788540/energy-consumptio...
In 2022, Google consumed 22.3G Watt-Hours of energy.
Total electricity consumption by humanity:
https://www.statista.com/statistics/280704/world-power-consu...
In 2022, it was 26T Watt-Hours.
Now, Google is a single company, and if we extrapolate with some cocktail-napkin math, let's say that similar tech giants put together consume, say, 20x Google? 50x Google? So between 2% and 6% of all human electricity consumption.
I realize that's not broken down for AI, but I'm sure if we do break it down we'll find that's an increasing fraction. In this article:
https://cse.engin.umich.edu/stories/power-hungry-ai-research...
the quoted figure is 2% of US electricity usage.
The only reason folks are only paying a small monthly subscription for gippity is literally because of all the VC money flowing in. Training, running, and scaling this stuff has a huge cost. The extra datacentres, the fresh water, all the air conditioning, energy usage, chips, the exploited unprotected labour, etc. It seems very expensive.
Usually when people selling shovels are giving shovels away for free they're banking on a payoff. Usually a regulatory capture payoff. Or a hedge of some kind?
I'm still trying to wrap my head around this.
Update: .. VC money flowing in and the extremely favourable taxation "exceptions" for tech companies in thirsty economies...
Is there a reason these articles like to say "nation-state" rather than "countries"? I think the question of whether China is a nation-state is not entirely settled (the state is broader than the nation in its case), but also it seems odd to exclude countries like Belgium that are not nation-states from these kinds of statements.
Unbiased AI is, I believe, an existential threat to the “powers that be” retaining control of the narrative, and must be avoided at all costs.
Related
Meta AI develops compact language model for mobile devices
Meta AI introduces MobileLLM, a compact language model challenging the need for large AI models. Optimized with under 1 billion parameters, it outperforms larger models by 2.7% to 4.3% on tasks. MobileLLM's innovations include model depth prioritization, embedding sharing, grouped-query attention, and weight-sharing techniques. The 350 million parameter version matches larger models' accuracy on specific tasks, hinting at compact models' potential for efficiency. While not publicly available, Meta has open-sourced the pre-training code, promoting research towards sustainable AI models for personal devices.
OpenAI slashes the cost of using its AI with a "mini" model
OpenAI launches GPT-4o mini, a cheaper model enhancing AI accessibility. Meta to release Llama 3. Market sees a mix of small and large models for cost-effective AI solutions.
Meta won't release its multimodal Llama AI model in the EU
Meta will not release its Llama AI model in the EU due to regulatory concerns, impacting European companies. Apple also considers excluding the EU from its AI rollout. This decision poses challenges for companies worldwide.
Llama 3.1: Our most capable models to date
Meta has launched Llama 3.1 405B, an advanced open-source AI model supporting diverse languages and extended context length. It introduces new features like Llama Guard 3 and aims to enhance AI applications with improved models and partnerships.
Meta releases an open-weights GPT-4-level AI model, Llama 3.1 405B
Meta has launched Llama 3.1 405B, a free AI language model with 405 billion parameters, challenging closed AI models. Users can download it for personal use, promoting open-source AI principles. Mark Zuckerberg endorses this move.