2023 was the year of Generative AI. The release of ChatGPT on November 30, 2022 catapulted us into a new world, one where terms like large language models and LLMs are now a household name. Work looks different today than it did in 2022.
As we enter a new year, what can we expect from generative AI?
Defining Moments of 2023
To answer this question, our co-founder Daniel Katz recently gave a presentation at the TLTF Summit in Miami where he provided a 2023 year in review - a year dominated by developments in Generative AI.
Building from that presentation, here are 13 defining moments from 2023 that collectively mark it as a watershed in generative AI -
3.30.23 - Verticalized LLMs Enter the Scene. BloombergGPT is released, a domain-specific LLM that was trained on 300+ billion finance tokens with the aim of performing financial NLP tasks. It outperforms similarly-sized open models on financial tasks and performs as well or better on general NLP benchmarks.
4.6.23 - Samsung Code Incident. A rise in FUD - fear, uncertainty and doubt - as folks (rightly) became concerned about the leakage of proprietary data.
5.27.23 - NY ChatGPT Lawyer Incident. The possibility of model hallucinations breeds more FUD.
6.26.23 - Casetext & Thomson Reuters Deal. The excitement and fear of missing out on generative AI increases.
7.18.23 - Llama 2 Release. Open-source LLMs are real competitors to OpenAI.
9.15.23 - Harvard Business School Study Release. Demonstrates an increase in productivity and quality of knowledge work when using LLMs.
11.6.23 - OpenAI Dev Day. A major focus on agents.
11.17.23 - Sam Altman Fired. OpenAI chaos highlights a need to move away from a single-source LLM strategy.
12.4.23 - TLTF Summit. The Legal Tech Fund Summit was a sort of “woodstock” for legal tech and innovation.
Here are some late 2023 developments that took place after the TLTF presentation -
12.6.23 - Google’s Gemini Release. A phone can run a small model of an LLM. The competition to OpenAI continues.
12.19.23 - Harvey raises 80M on 715M Valuation. A reported 70x multiple on revenue and is a huge bet by investors. Substantial investor confidence in the potential of generative AI but a pretty large revenue multiple.
12.31.23 - Chief Justice Roberts Year End Report Addresses AI. Generative AI has gone mainstream.
Forecasts for 2024
Drawing on insights from these events of 2023, we can anticipate with a measured degree of confidence the following five forecasts for 2024 -
Private Instances of LLMs. Partly in response to FUD around infosec and partly to increase capabilities, the new norm will be private instances of LLMs - i.e., firms invoking their own data within an LLM.
The K-LLM Strategy. Adopters will mix and match LLMs rather than rely on a single LLM to do everything (i.e., the single-sourced LLM strategy).
Rise of Agents. More agents - software that uses LLMs to plan and execute multi-step tasks - will be developed to be tunable and configurable to specific workstreams of organizations.
Verticalized LLMs. There will be an acceleration of verticalized large language models. Rather than simply rely on generalized large language models, it’s becoming increasingly evident that domain-specific LLMs have a significant role to play.
Upskilling on Generative AI. To meet the demands of utilizing new technology, organizations will have significant incentive to upskill their people on generative AI.
The profound impact of Generative AI is irrefutable. Our global landscape has rapidly transformed and will continue to do so. The remaining question is whether individuals will willingly embrace and adapt to this change or wait until the choice becomes a necessity.