Meta’s new AI models are making waves in tech circles. The two new models, part of parent company Facebook’s Llama line of artificial intelligence tools, are both open source, helping them stand out from competing offerings from OpenAI and other well-known names. known.
Meta’s new Llama models have underlying data sets of varying sizes, with the Llama 3 8B model having eight billion parameters and the Llama 3 70B model having some seventy billion parameters. The more parameters, the more powerful the model, but not all AI tasks require the largest possible data set.
The company’s new models, which were trained on 24,000 GPU clusters, performed well in the benchmarks Meta put them against, outperforming some competitors’ models already in the market. What matters to those of us who aren’t competing to build and release the best or biggest AI models, what we care about is that they get even better over time . And the work. And one plot Calculation.
While Meta takes an open source approach to AI work, its competitors often prefer more closed work. OpenAI, despite its name and its history, offers access to its models, but not to their source code. There is a healthy debate in the AI world regarding the best approach, both in terms of speed of development and security. After all, some technologists – and some IT specialists, to be clear – worry that AI technology is developing too quickly and could prove dangerous for democracies and much more.
For now, Meta keeps the AI fire burning, providing a new challenge to its peers and rivals to improve their latest releases. Press play and let’s talk!
techcrunch
