In the world of AI, the floodgates seem to have opened, inundating us with a deluge of new AI models practically every week. The sheer volume, averaging around 10 releases recently, begs the question: how many AI models are too many?
Currently AI is undeniably in a state of flux, characterized by a proliferation of models from both niche developers and well-established entities with hefty funding. This week alone, we witnessed the debut of several noteworthy models, each vying for attention in an increasingly crowded field.
Here’s a snapshot of some of the latest AI models arrivals this week:
- LLaMa-3: Meta’s latest offering, touted as an “open” flagship large language model, albeit amid some controversy over its openness.
- Mistral 8×22: A sizable ‘mixture of experts’ model from a French entity that has somewhat retreated from its previous embrace of openness.
- Stable Diffusion 3 Turbo: An upgraded iteration of SD3, aligned with the newly introduced open-ish Stability API.
- Adobe Acrobat AI Assistant: Primarily positioned as a conduit for interacting with documents, likely leveraging ChatGPT under the hood.
- Reka Core: Engineered from scratch by a former Big AI team, challenging established models with its multimodal capabilities.
- Idefics2: A more open multimodal model building on the foundations laid by Mistral and Google.
- OLMo-1.7-7B: A larger variant of AI2’s LLM, aiming for openness and scalability.
- Pile-T5: A refined version of the trusty T5 model fine-tuned on code repositories from the Pile dataset.
- Cohere Compass: An ’embedding model’ focused on integrating diverse data types to cater to a wide array of use cases.
- Imagine Flash: Meta’s latest entrant in the image generation domain, leveraging novel distillation techniques for enhanced diffusion.
- Limitless: An intriguing concept of personalized AI spanning various platforms, promising an omnipresent digital assistant.
While this roster may seem exhaustive, it’s just the tip of the iceberg. The reality is, the AI world is evolving at breakneck speed, with numerous models and tools emerging each week, making it virtually impossible to keep pace with every development.
However, amidst this whirlwind of innovation, a fundamental shift is underway. Models like ChatGPT and Gemini have evolved beyond mere models, evolving into comprehensive platforms catering to diverse use cases. Conversely, models like LLaMa and OLMo operate more discreetly, serving as backend components rather than standalone entities.
While it’s impractical to track every model, we endeavor to highlight the most significant advancements, offering insights into the developments shaping the AI sector.