The following are some of the news items mentioned in the last few days:

  • AI is poised to outperform humans in writing code as leading groups including OpenAI, Anthropic and Google race to release systems that are reshaping the software industry
    • Open AI announced Codex CLI, an AI agent designed to use its models to help users with coding tasks
    • In 2023, AI systems were only able to solve 4.4 percent of coding problems based on an industry test called SWE-bench. That figure jumped too 69.1 percent this year
      • The SWE-bench (Software Engineering Benchmark) is a benchmarking framework used to assess the performance of AI systems in solving software engineering tasks, particularly coding problems. The test evaluates AI models on their ability to complete programming-related challenges, such as solving algorithmic problems, debugging code, or writing efficient code solutions.
      • This significant improvement from 4.4% to 69.1% suggests that AI systems have made remarkable progress in their ability to handle coding problems and demonstrate practical competence in software engineering tasks. This improvement could be attributed to advancements in AI models, better training data, and improvements in model architecture (like larger and more capable language models such as GPT-4)
  • UAE deploys AI to help write laws, despite reliability fears
  • Sam Altman said last week that its audience had doubled in a matter of weeks and now comprises a tenth of the worlds population.
    • Revenue jumped to 4B USD and it currently valued at 300B USD. Google didn’t reach that valuation until its annual revenue reached 60B USD
    • It took Google 13 years to reach 1 billion users milestone and Facebook 8 years. OpenAI looks to get to 1 billion users in 3 years