Large language models- Brace for shakeout in gen AI world

Large language models: Brace for shakeout in gen AI world

By Jaspreet Bindra

I believe that 2024 will be the real ‘Year of AI and GenAI’, putting even 2023 in the shade.The AI race accelerates: Google’s Gemini has put further fuel in the GenAI fire, and there will be a race between OpenAI-Microsoft, Google, Meta and well-funded startups like Inflection and Anthropic to build newer, more powerful GenAI models. We might see GPT4.5 or 5, and Gemini Ultra will be launched. However, there will also be a shakeout, with stability being perhaps the first casualty.LLMs become LMMs: Large language models will become large multimodal models, going beyond text to pictures, videos, sound, etc., and some of them will have agency and power to perform tasks on our behalf. We will also see SLMs or ‘small language models’ built for specific tasks with less power, like Gemini Nano and Microsoft’s Orca. There will be a race to AI regulation: As the concerns around AI increase, so will the focus on regulation. The EU AI Act will be the first one to be made law, other countries will follow suit, and India will come up with a light-touch AI regulation of some kind.

GenAI will transform work: Microsoft CoPilot-like products will have as great an impact on our work and productivity as PCs, GUI (Graphical User Interface) or productivity software like Excel and Word had a few decades back. Co-pilots will get into a higher gear next year. This will further have an impact on jobs by AI, and some companies will ‘use’ AI as a shield to do efficiency layoffs! Proprietary and open-source GenAI will coexist: Much like in software, the GenAI LLM world is bifurcating into proprietary LLMs by companies like OpenAI and Google, and open-source ones by Meta and Mistral. Proprietary gives the benefits of safety and control, while open source decentralise and democratise. Big players will do both.

AI deepfakes will disrupt: There are elections in the biggest democracies in the world, including India, and some political entities will use deepfakes to influence the narrative and electoral process. This will prompt  stricter regulation and education and awareness campaigns around deepfakes. AI and GenAI are here to stay. AI is just not a technology, but as Gartner puts it, a fundamental shift in how humans and machines interact. Humanity will need to figure out how to leverage the astonishing benefits that AI can bring to us, and also manage some of the potential harms. That effort will start in earnest in 2024.

The author is MD, founder, The Tech Whisperer, a UK-based firm

Follow us on Twitter, Facebook, LinkedIn