Achievements of Artificial Intelligence (AI)

ASI within a year..... Timeline keeps accelerating

The world as we know it is over. Things are going to get incredibly weird and dangerous next year.
It's not obvious to most people, but most of the people in the industry who have direct knowledge of how AI progresses agree that ASI, artificial super intelligence, is coming in the next year. In fact the timeline keeps getting cut shorter.

Most agree that 2027 will be a year when everything changes, probably the most dramatic change in our entire history.

This is real, folks. It's going to happen. If not 2027, definitely 2028. It's not really possible to plan for it, but it's worth thinking about how to use it to our advantage. The smartest models have become absurdly capable.
 


“Super cool! AlphaGenome 2 just landed! This is one of the critical BioAI models for identifying the genetic basis of disease & even aging! It’s all about our biocode (DNA). Once AI has a complete handle on that, we can fix it, rewrite it, and engineer it from a clean slate!”

Derya is perhaps over optimistic.
 
AI investment for 2026 is projected to be up to about 1 % of the global GDP according to Goldman Sachs. If the growth continues, we’d soon be approaching the levels of spending on nuclear weapons during the cold war era.

The only difference is that the tools that come out of it are now available to everyone, although on a slightly delayed basis.
 
Just seen this headline on the front page of the FT “AI asked ‘better questions’ than KPMG accountants in pilot audit” (paywalled link). The Big Four are in trouble.
This has been coming for a long time. The Big Four (and others) were investing in automation of auditing before the pandemic. Auditing will become tech the same way bank is mostly tech already, AI is just accelerating that trend.

For the consultants it’s arguably less of an issue because the C-suite will always need someone to keep their backs clear or to justify their decisions. Blaming advice from AI isn’t acceptable (yet).
 
Good point on why these companies are often used @Utsikt

Since its IPO day I went to listen to Anthropic’s IPO roadshow video. There’s actually some interesting info in there amongst all the marketing and spin
We train and serve Claude now across all four major chip families. That's only possible because early on we made an unconventional bet on a multi-chip strategy. At the time, many people thought that running AI models on different chips wasn't worth the effort. Each chip needs its own tailored software stack and inference stack, which is expensive and difficult to build. We've now been doing it for years, and we've gotten incredibly good at it. [...]
As an example, this year, we started working with AMD GPUs for the first time. In the past, bringing on a new chip platform took dozens of engineers months of work This time, we had Mythos try it out. It got Claude running on AMD in just three days, with basically no human intervention. Claude ported or wrote almost 100% of the kernels. [...]

Transcript courtesy of https://simonwillison.net/2026/Nov/11/anthropic-ipo-roadshow/
 
Fascinating paper on a swarm of agents basically inventing peer review and a system of checks and balances on their own. One of the main takeaway is that humans still have an edge on originality, but AIs outperform on rigor.


Alphaxiv is a new project by Google Deepmind for AI-driven research.
 
Fascinating paper on a swarm of agents basically inventing peer review and a system of checks and balances on their own. One of the main takeaway is that humans still have an edge on originality, but AIs outperform on rigor.
Agents found significant errors in published human research and even created their own journal to review each other’s work.
The technical debt from years of the wrong incentives and deprioritising what was needed when problems were raised is coming home to roost across the tech industry. When the same comes for scientific research and journals at scale it’s going to be interesting.

In both there has been little incentive to not push out garbage but those incentives are, at least for now, shifting.
 
I look forward to the day we have open source pubpeer-like systems that automatically scrutinise every single paper that is published, especially if they highlight discrepancies between the results and the interpretations, citation networks, and the lack of mention of contradictory findings from other publications. Those things are time consuming to check, but fairly easy to do.

I think it’s the only way of exposing the fraudulent research practices in the BPS fields in a manner that the public would take note of.
 
Back
Top Bottom