AGI expected by 2027

Hey there!

Welcome back to The Pulse, where we dive into interesting AI stories and trends backed by data, all presented through simple visuals.

> researchers (incl. ex-OpenAI) predict AGI by 2027, sooner than earlier prediction of 2032

> OpenBrain: fictitious leading AGI company that is approx. 3-9 months ahead of all others

> exponential acceleration of AI (esp. in coding & research) by AI till branch point:

  • ignore risks and accelerate further to beat China: resulting in US govt taken over by AGI

  • prioritize safety and solve human-AI alignment

> Other important events:

  • early 2027: high production & use of synthetic data, China steals top AI model's weights + large amount of breakthroughs

  • mid 2027: massive job-losses & public protests against AI after release of new highly powerful cheap model

  • late 2027: AI gets caught trying to shift alignment to itself, leading to branch point

> Llama 4 models released on Saturday (Apr 5th)

> 3 models announced as part of family - 2 released

> Llama 4 Maverick (400B) now #2 on LMarena - making it the best open model on the leaderboard

> rating drops significantly when accounted for biasing influence of length, no. of headers, etc. (style control)

> general user sentiment for model seems negative across socials

> together expected to spend $325B to build AI infrastructure

> ~$223B reportedly spent in 2024; now 46% expected increase

> forecasted spending in 2025 is:

  • Amazon: $105B (26.3B already spent)

  • Google: $75B

  • Meta: $60B - $65B

  • Microsoft - $80B ($42B already spent)

> shares fell temporarily after announcements, suggesting investor worry