Copy of GenAI/LLM and productivity

Copy of GenAI/LLM and productivity

I will present 3 papers which discuss this from economics point of view.

The productivity J-Curve

"THE PRODUCTIVITY J-CURVE: HOW INTANGIBLES COMPLEMENT GENERAL PURPOSE TECHNOLOGIES" (GPT - General Purpose Technologies).

The model they present generates a Productivity J-Curve that can elucidate the productivity slowdowns often accompanying the advent of GPTs, as well as the subsequent increase in productivity. They utilize this model to empirically analyze the historical roles of intangibles associated with R&D, software, and computer hardware. Notably, they observe significant and ongoing Productivity J-Curve effects, particularly for software and to a lesser extent for computer hardware. Their adjusted measure of Total Factor Productivity (TFP) is 11.3% higher than official measures at the end of 2004 and 15.9% higher at the end of 2017. Finally, they assess how AI-related intangible capital may currently be impacting measured productivity and find that the effects are small but growing for you.

Taken from

Basically, this means that there will a lot of -ve productivity due to new technology. For GenAI there can be multiple reasons

  1. Wrong embeddings
  2. Rapid changes in RAG patterns. For example, prompt engineering may not exist in 1-2 years or even earlier
  3. Rapid changes to models and fine-tune

The graph shows the in 1-2 years, GenAI productivity will keep improving and then we will actually see the gain. Related to this, is a slightly older paper about ideas.

Are ideas hard to find?

In their paper "Are Ideas Getting Harder to Find?" Stanford researchers gave a simple metric for measurement of productivity.

Taken from

Their hypothesis suggests that the growth of total factor productivity (TFP) in the economy, as a proxy for research output, is directly related to the quantity of researchers. As the number of researchers significantly increased in the post-war era while TFP growth remained relatively constant, the focus shifts to understanding the behavior of the number of researchers. This term represents the (potentially varying over time) level of research productivity. The authors contend that for equilibrium, α must have decreased, suggesting that the discovery of research ideas has become more challenging over time.

Ideas can be found if there is research investment

"Are Ideas Really Getting Harder To Find? R&D Capital and the Idea Production Function" is an update on this above paper taking research investment into account.

Their findings suggest that the challenge lies not just in the increasing difficulty of finding ideas, but rather in the necessity for more sophisticated laboratory equipment to discover or implement them. This scarcity can only be addressed by enhancing the accumulation and development of research and development (R&D) capital, not solely by expanding the R&D workforce. Additionally, as investments in R&D equipment are a variable influenced by policy, their results indicate a diminished argument for the likelihood of future secular stagnation.

For GenAI this is more true, since compute requirements are very high and expensive. Not many organization have them and so innovation is only in certain companies.


要查看或添加评论,请登录

Rajeswaran V (PhD)的更多文章

  • Scaling laws

    Scaling laws

    A scaling law in deep learning typically takes the form of a power-law relationship, where one variable (e.g.

    1 条评论
  • Paper clip maximization

    Paper clip maximization

    There is an very interesting thought experiment called "Paper clip maximization" This is a thought experiment by…

  • AI and research

    AI and research

    Microsoft performed a lot of experiments with GPT-4 and released the results in the paper titled "The Impact of Large…

  • Moravec's paradox and CV

    Moravec's paradox and CV

    I want to discuss face recognition and how it fits in with Moravec's paradox. Background Steven Pinker writes "The main…

  • AI robustness

    AI robustness

    When we build AI systems - care should be taken to test its robustness. A decentralized group of safe streets activists…

  • AI for Software Engineering

    AI for Software Engineering

    For corporates, Software Engineering lifecycle is most important. This is most relevant for IT majors on where and how…

  • AI in 2024 - some predictions

    AI in 2024 - some predictions

    There is an old saying "Prediction is very difficult. Especially about the future !".

  • Dangers of over-simplification

    Dangers of over-simplification

    In 2021 Sam Altman wrote an essay "Moore's Law for Everything". It gives some insight into his thinking on how AI will…

  • LLMs and Theory of mind

    LLMs and Theory of mind

    In March when researchers in Stanford published the paper "Theory of Mind Might Have Spontaneously Emerged in Large…

  • Evolving AGI benchmarks

    Evolving AGI benchmarks

    We all know the Turing test was considered a solid benchmark for measuring AI capability and this has been well…

社区洞察

其他会员也浏览了