2023 is going to be weird. The tech industry is in cost-cutting mode, companies are doing layoffs, and VC funding just evaporated. And then ChatGPT shipped and everyone realized AI actually works now. So you've got two forces in conflict: "spend less money" and "our AI budget needs to ship yesterday". That tension will define the year. Here's what I think plays out.
The AI inflection point
GPT-3.5, the model underlying ChatGPT, crossed a quality threshold in late 2022. The combination of instruction following and language fluency is qualitatively different from previous LLM outputs. This arrived at a moment when most enterprises were cutting AI budgets in response to 2022's market conditions. The tension between pressure to reduce AI spending and the availability of AI capability that was newly useful will define the year.
Cloud cost optimisation
The zero-interest-rate environment of 2020-2022 made cloud cost optimisation a secondary concern for growth-stage companies. The 2022-2023 market correction reversed this. Engineering teams that had provisioned infrastructure for anticipated growth are rightsizing. FinOps, the practice of applying financial operations principles to cloud spend, moved from a nice-to-have to a function with dedicated headcount. Reserved instance purchases, savings plans, and spot instance architectures became priorities.
The talent market correction
The software engineering hiring boom of 2020-2021 reversed in 2022-2023. Major technology companies announced large layoffs. The labour market for software engineers shifted from candidate-driven to employer-driven for the first time since 2013. Engineering teams that had grown rapidly were being asked to do more with fewer people. This is a forcing function for productivity tools, including AI coding assistants.
What to watch in 2023
The specific things to track: the enterprise adoption rate of ChatGPT and GPT-4 for production applications, whether Bing's AI integration moves meaningful search share from Google, the open source model capability gap closing as Llama and its derivatives improve, and the regulatory response to generative AI's intellectual property questions. 2023 will produce answers to questions that 2022 only raised.