|
Anthropic Warns About AI Acceleration
Leading AI firm Anthropic, creator of Claude, has issued a new report titled When AI Builds Itself, in which it explains that, “the rate at which AI models improve is accelerating”, alongside some impressive numbers detailing how task completion capacity is now doubling every four months, which is up from every seven months previously.
Anthropic also predicts that on current trends, “in 2027, AI systems could be capable of tasks that take a person weeks.”
But what really caught attention were two key points:
● A potential scenario in which “AI systems themselves become capable of full recursive self-improvement, and begin building their successors.”
● A suggested pause in development, with Anthropic stating,
“We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology.”
It’s very unusual to hear a tech company (or any other kind of company) state that progress is going too well, so that second statement certainly stands out.  On the other hand though, a cynic might point out that Anthropic is lining up one of the largest IPOs the world has ever seen, in direct competition with both OpenAI and SpaceX (which includes xAI), and so making ominous statements works in their favor by keeping focus on the company and its tech.
No Fast Track Index Inclusion for SpaceX
Sticking with the subject of IPOs, a statement from the S&P Dow Jones Indices has clarified that there will be no changes to the criteria for S&P 500 inclusion with regard to megacap corporations.
This means that SpaceX will not be fast-tracked for index inclusion, in contrast to recent speculation that the rules would be changed.  There had been a lot of discussion around fast-track inclusion, with some passive investors framing it as being forced to buy SpaceX at an inflated valuation, so for those opposed, this is good news.
The upshot then, is that SpaceX will have to trade publicly for at least 12 months before inclusion (rather than a proposed six months), and meet existing criteria around profitability, float, liquidity, and market cap.
Zcash Crashes on Bug Discovery
Crypto ended the week with another catastrophe, as Zcash revealed a vulnerability that immediately caused the ZEC price to crash by over 50%.
In short, the sequence of events was:
● Zcash, using AI tools, found a critical bug in Orchard, a shielded pool.
● That bug has been around for four years, and could potentially have allowed the creation of unauthorized ZEC, meaning the supply may have been inflated.
● Because Orchard is private, it’s not possible to check whether any fake tokens were created.
● Developers fixed the bug and the Zcash Foundation states there is currently no evidence of unauthorized token creation.
● A further upgrade has been proposed in order to verify that no tokens were created.
So, although the bug may not have been exploited, this cannot–through the nature of the privacy protocol–be immediately verified for certain.
And the longer-term issue is that the four-year-old bug was only now uncovered using AI tools. This raises the likelihood of there being more vulnerabilities across other protocols, although keep in mind that AI discovery tools are available to security teams as well as attackers, so we’re really just witnessing the next phase of an endless cyber-security battle.
| 
