Looking back at the end of 2022 (November) when the bomb called ChatGPT was dropped by OpenAI, with hindsight, the reverberation wave from that impact has overcome all frictional resistances with no sign of slowing down. Acceleration seems to be the reading from everyone's speedometer.
2023 has undoubtedly been the year of generative AI. Powerful new models from Open AI like GPT4, DALL-E 3, and others like Stable Diffusion have captured the public imagination and shown glimpses of how AI can automate creative and knowledge work. As we approach 2024, generative AI promises to become even more capable and mainstream. Here's an overview of the generative AI revolution in 2023 and key predictions for where it goes next.
The Explosion of Foundation Models:
In 2023, the trend was undoubtedly towards the development of chatbots, and crafting user interfaces powered by extensive large language models. This marked the year when LLM bots evolved from purely text-based interactions to a more versatile, multimodal functionality.
Here, we saw an expansion of what AI researchers call "foundation models" - versatile, large language models that can be adapted to various tasks. Open AI was at the forefront again with GPT4. Anthropic introduced Claude, displaying impressively robust conversational ability. Google opened up LaMDA. Meta launched BlenderBot. These models showed the rapid progress of conversational AI.
Meanwhile, image generators like DALL-E 3 and Stable Diffusion harnessed a different modality to generate striking images from text prompts. LLMs like Anthropic's Claude, Meta's Llama, Google's PaLM, and Inflection AI all carved out their lane in the race, with Hugging Face being the GitHub for Gen AI. The advances this year in both language and image generation foreshadow the emergence of multimodal foundation models that manipulate images, text, code, and potentially even video and audio with ease. It was the year the tech behemoths took the reins and established control of the landscape. Microsoft and Google integrated LLMs into search and enterprise solutions, while Amazon championed a model-as-a-service approach. Meta took a different route, carrying the torch for the open-source community.