Club Benefits
You are now subscribed
Your newsletter sign-up was successful
Want to add more newsletters?
Daily (Mon-Sun)
Tom's Guide Daily
Sign up to get the latest updates on all of your favorite content! From cutting-edge tech news and the hottest streaming buzz to unbeatable deals on the best products and in-depth reviews, we’ve got you covered.
Weekly on Thursday
Tom's AI Guide
Be AI savvy with your weekly newsletter summing up all the biggest AI news you need to know. Plus, analysis from our AI editor and tips on how to use the latest AI tools!
Weekly on Friday
Tom's iGuide
Unlock the vast world of Apple news straight to your inbox. With coverage on everything from exciting product launches to essential software updates, this is your go-to source for the latest updates on all the best Apple content.
Weekly on Monday
Tom's Streaming Guide
Our weekly newsletter is expertly crafted to immerse you in the world of streaming. Stay updated on the latest releases and our top recommendations across your favorite streaming platforms.
Join the club
Get full access to premium articles, exclusive features and a growing list of member rewards.
Google DeepMind has just dropped a brand new AI model, Gemma 2 2B. It’s a 2 billion parameter model that is small enough to fit onto a smartphone and still offers GPT 3.5 levels of performance.
The launch follows on from the June announcement of the Gemma 2 9B and 27B models, all of which represent Google's fightback against rivals in the AI LLM space, particularly against Meta and its Llama 3.1 family.
According to Google the new model offers ‘best-in-class performance for its size, outperforming other open models in its category,’ and the LYMSYS charts certainly show some impressive stats.
The model apparently achieves this performance uplift ‘by learning from larger models through distillation.’ The technique was introduced in 2015 in a paper by Geoffrey Hinton, and employs a sophisticated student-teacher method of training that allows for smaller, more compressed models.
So for most people, cloud-based Al is likely to continue to dominate, at least in the short term. But inevitably, as the technology matures, on-device AI is going to take a larger slice of the chatbot pie.
The market for mobile and edge-based Al, for use on small portable and static devices such as sensors, is expected to grow exponentially over the coming years. This is especially true of multimodal AI applications which will use vision and sound to create a rich interactive universe for the user. Hence the race to perfect smaller, more powerful systems as quickly as possible.
However, despite all the talk of running this type of AI locally on smartphones (aka ‘on-device’), the reality is there are still very few portable devices that can cope with the processing demands of current LLMs.
Modern iPhones running at least IOS 14, and later Android phones using Snapdragon processors like the Galaxy S23 Ultra, are some that can. But even then the performance is modest in comparison to computer-based systems and Apple's Apple Intelligence which is mostly on-device requires at least an iPhone 15.
So for most people, cloud-based Al is likely to continue to dominate, at least in the short term. But inevitably, as the technology matures, on-device AI is going to take a larger slice of the chatbot pie. Especially as users begin to focus more and more on privacy, security and reliability.
Interested readers can play around with the new Gemma 2 2B model at the Google AI Studio playground and it is available through Ollama to run on device.
More from Tom's Guide
- 5 astonishing uses of AI happening right now
- Sam Altman hopes to take on Nvidia with new global network of AI chip factories
- AMD drops new bargain-minded CPUs and GPU with an AI focus at CES 2024
