Skip to content

Instantly share code, notes, and snippets.

@dfoxfranke
Created March 17, 2023 19:39
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save dfoxfranke/b0e8dffb6fbaa29ac67667ad4d555d71 to your computer and use it in GitHub Desktop.
Save dfoxfranke/b0e8dffb6fbaa29ac67667ad4d555d71 to your computer and use it in GitHub Desktop.
5-year predictions about the future of artificial intelligence

5-year predictions about the future of artificial intelligence

Written in March 2023, predicting the state of the world in March 2028. Percentages denote confidence levels.

Broadly

AI will be an enormous contributor to economic growth over the next five years, as important as the Internet was to the '00s. However, it will continue to be the case that AI's ability profile is spikey and inconsistent, sometimes performing with vastly superhuman ability and other times falling flat on its face in ways that no healthy human ever would. Getting past this will take much more than five years (perhaps twenty) and require tens more innovations each as fundamental and important as the invention of LLMs, and until that happens, AI will not have any "singularity"-scale consequences. For the next five years, AI will remain a powerful tool wielded by humans rather than becoming a master of humanity's fate.

Extreme changes to the human condition

  1. Humans will continue to exist as biological entities who are born, reproduce, and die, and have thoughts and emotions recognizable to us today (98%).

  2. Human-AI interaction will continue to be through our normal sensorium rather than directly augmenting our neurology. Exceptions may exist (e.g. for patients with locked-in syndrome) but they will not be salient to most people's everyday life. (95%)

Economics

  1. World GDP will be more than 130% of what it is today (70%).

  2. World GDP will be less than 300% what it is today (95%).

  3. The Gini Index among residents of the land area that today makes up the G7 will be less than 70%. (90%)

Writing

  1. At least a million people will have read a novel (not necessarily all the same one) written entirely by an AI working from a brief (<500 word) prompt with no human postprocessing of the manuscript. (70%)

  2. At least 90% of human time spent reading novels will be on ones whose production involved substantial human effort. (80%)

  3. Most newly-published long-form writing will incorporate content or major edits which were suggested by an AI. (80%)

  4. Most people who today earn their living as developmental editors will no longer be able to do so. (70%)

Visual Art

  1. The market for human-produced fine art will be at least 75% as large as is it is today. (90%)

  2. Animation studios will employ at least half as many artistic professionals as they do today. (80%)

  3. Most people who today earn their living as book illustrators will no longer be able to do so. (70%)

Transportation

  1. Autonomous vehicles, capable of operating in favorable conditions and pulling over to a safe stop when they become unfavorable, will be available to a market of at least 25 million consumers. (60%)

  2. Autonomous vehicles will not be good enough that people who live in snowy rural areas can skip learning to drive. (80%)

  3. Commercial flights will still have human pilots present in the cockpit, even if they don't do much. (90%)

  4. Most people who today earn their living as long-haul truckers will no longer be able to do so. (70%)

Software

  1. At least 90% as many people as today will be employed as software engineers. (80%)

  2. Nearly everyone employed as a software engineer will routinely use AI to generate significant pieces of code. (90%)

  3. No F500 company will have any business-critical software written entirely by AI without human review and modification. (90%)

  4. No profitable company will exist which is founded and run entirely by an AI. (95%)

Education and child care

  1. AI tutoring will have a major role in all levels of education. (90%)

  2. It will a reasonable, but not necessarily mainstream, option for a child to get an entire K-12 education from home by learning from an AI, and parents providing only minimal supervision. (70%)

  3. It will not be possible to completely entrust all aspects of child-rearing to a robot nanny. (95%)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment