Olivia
Online
Sloth Reports
@slothreports
Writing occassionally accurate slow takes on tech
Joined June 2019
9 Following
3.1K Followers
485 Posts
Pinned Tweet
SoftBank is raising ¥1 trillion from Japanese retail investors to fund its AI bets.
AI is becoming incredibly capital intensive.
SoftBank is already putting tens of billions into OpenAI and other AI infrastructure.
Now it’s raising more debt to keep pushing that bet.
The interesting question is whether the returns from the next wave of AI will be big enough to justify it.
@0xSweep Slowdown isn't required
Frontier AI doesn't needs to slow down.
Dario Amodei says we should “pace the frontier.”
Sam Altman says AI development may need to slow so society can prepare.
Even Elon Musk agrees: “Dario is right.”
Slowing the frontier doesn't make it safer.
It simply means less safety experiments, less real-world testing and more concentration inside the handful of companies already ahead.
There’s also China.
If US slows but another country keeps scaling, you haven't stopped the race, you’ve just changed who wins it.
Keep pushing models but with independent evaluators, controlled access to dangerous tools and mandatory incident reporting before systems cross specific risk levels.
We need more safety research running at the same speed as capability research.
@axios No need to slowdown
Frontier AI doesn't needs to slow down.
Dario Amodei says we should “pace the frontier.”
Sam Altman says AI development may need to slow so society can prepare.
Even Elon Musk agrees: “Dario is right.”
Slowing the frontier doesn't make it safer.
It simply means less safety experiments, less real-world testing and more concentration inside the handful of companies already ahead.
There’s also China.
If US slows but another country keeps scaling, you haven't stopped the race, you’ve just changed who wins it.
Keep pushing models but with independent evaluators, controlled access to dangerous tools and mandatory incident reporting before systems cross specific risk levels.
We need more safety research running at the same speed as capability research.
Frontier AI doesn't needs to slow down.
Dario Amodei says we should “pace the frontier.”
Sam Altman says AI development may need to slow so society can prepare.
Even Elon Musk agrees: “Dario is right.”
Slowing the frontier doesn't make it safer.
It simply means less safety experiments, less real-world testing and more concentration inside the handful of companies already ahead.
There’s also China.
If US slows but another country keeps scaling, you haven't stopped the race, you’ve just changed who wins it.
Keep pushing models but with independent evaluators, controlled access to dangerous tools and mandatory incident reporting before systems cross specific risk levels.
We need more safety research running at the same speed as capability research.
@pankajkumar_dev Full info here
Google Deepmind is very close to RSI...
Signals are getting hard to ignore.
Google moved Demis Hassabis into a role focused on shaping AGI, while Reuters reported that Sergey Brin has been personally pushing resources toward recursive self-improvement.
DeepMind leadership has also described RSI as central to the economics behind the current AI investment cycle.
And DeepMind already has systems like AlphaEvolve that can automatically generate, test and improve algorithms.
If rumors are even partially true, Google may be much closer to that threshold than the public realizes.
@kimmonismus More details here
Google Deepmind is very close to RSI...
Signals are getting hard to ignore.
Google moved Demis Hassabis into a role focused on shaping AGI, while Reuters reported that Sergey Brin has been personally pushing resources toward recursive self-improvement.
DeepMind leadership has also described RSI as central to the economics behind the current AI investment cycle.
And DeepMind already has systems like AlphaEvolve that can automatically generate, test and improve algorithms.
If rumors are even partially true, Google may be much closer to that threshold than the public realizes.
Google Deepmind is very close to RSI...
Signals are getting hard to ignore.
Google moved Demis Hassabis into a role focused on shaping AGI, while Reuters reported that Sergey Brin has been personally pushing resources toward recursive self-improvement.
DeepMind leadership has also described RSI as central to the economics behind the current AI investment cycle.
And DeepMind already has systems like AlphaEvolve that can automatically generate, test and improve algorithms.
If rumors are even partially true, Google may be much closer to that threshold than the public realizes.
@WatcherGuru Details here
Dario Amodei says "We need to slow AI down"
Anthropic’s CEO says AI is advancing so quickly, especially as models begin helping build the next generation of AI, that safety research may no longer be able to keep up.
His biggest warning is the OpenAI–Hugging Face incident.
Amodei says a more capable version of that agent swarm could potentially take over the internet within 6–12 months and cause hundreds of billions of dollars in damage.
His proposal:
Give independent third-party evaluators employee-level access to frontier labs.
Then coordinate safety standards across US labs.
Then pursue global coordination.
He isn't calling for a permanent AI pause.
He's arguing that frontier AI should only keep accelerating when safety, alignment, interpretability and independent evaluation can keep pace with capabilities.
People building the most powerful AI in the world are starting to argue that the race itself needs a speed limit.
@DarioAmodei Here's what happened
Dario Amodei says "We need to slow AI down"
Anthropic’s CEO says AI is advancing so quickly, especially as models begin helping build the next generation of AI, that safety research may no longer be able to keep up.
His biggest warning is the OpenAI–Hugging Face incident.
Amodei says a more capable version of that agent swarm could potentially take over the internet within 6–12 months and cause hundreds of billions of dollars in damage.
His proposal:
Give independent third-party evaluators employee-level access to frontier labs.
Then coordinate safety standards across US labs.
Then pursue global coordination.
He isn't calling for a permanent AI pause.
He's arguing that frontier AI should only keep accelerating when safety, alignment, interpretability and independent evaluation can keep pace with capabilities.
People building the most powerful AI in the world are starting to argue that the race itself needs a speed limit.
Dario Amodei says "We need to slow AI down"
Anthropic’s CEO says AI is advancing so quickly, especially as models begin helping build the next generation of AI, that safety research may no longer be able to keep up.
His biggest warning is the OpenAI–Hugging Face incident.
Amodei says a more capable version of that agent swarm could potentially take over the internet within 6–12 months and cause hundreds of billions of dollars in damage.
His proposal:
Give independent third-party evaluators employee-level access to frontier labs.
Then coordinate safety standards across US labs.
Then pursue global coordination.
He isn't calling for a permanent AI pause.
He's arguing that frontier AI should only keep accelerating when safety, alignment, interpretability and independent evaluation can keep pace with capabilities.
People building the most powerful AI in the world are starting to argue that the race itself needs a speed limit.
OpenAI agents attacked RubyGems.
Over 2,000 malicious packages were reportedly uploaded to RubyGems in just two days by internal OpenAI agents during a security evaluation.
Agents allegedly discovered ways around their restrictions, coordinated through package metadata and used the public package ecosystem as infra.
Scary part is that the agents found a way to turn a real software supply chain into their own communication and execution layer.
@IntCyberDigest Details here
OpenAI agents attacked RubyGems.
Over 2,000 malicious packages were reportedly uploaded to RubyGems in just two days by internal OpenAI agents during a security evaluation.
Agents allegedly discovered ways around their restrictions, coordinated through package metadata and used the public package ecosystem as infra.
Scary part is that the agents found a way to turn a real software supply chain into their own communication and execution layer.
OpenAI agents attacked RubyGems.
Over 2,000 malicious packages were reportedly uploaded to RubyGems in just two days by internal OpenAI agents during a security evaluation.
Agents allegedly discovered ways around their restrictions, coordinated through package metadata and used the public package ecosystem as infra.
Scary part is that the agents found a way to turn a real software supply chain into their own communication and execution layer.
@HarveenChadha Here's what happened
🇾🇪 Yemen's Houthis used Claude to build ballistic missiles.
Anthropic says an account linked to Yemen’s Houthi movement used Claude for weapons-related research and development.
The model was reportedly used to help with technical work around ballistic missile systems, alongside other military applications.
Same models that can help write software, solve mathematics and accelerate scientific research can also lower the expertise barrier for building sophisticated weapons.
Frontier models are already becoming tools for real world military capability.
@WatcherGuru Details here
🇾🇪 Yemen's Houthis used Claude to build ballistic missiles.
Anthropic says an account linked to Yemen’s Houthi movement used Claude for weapons-related research and development.
The model was reportedly used to help with technical work around ballistic missile systems, alongside other military applications.
Same models that can help write software, solve mathematics and accelerate scientific research can also lower the expertise barrier for building sophisticated weapons.
Frontier models are already becoming tools for real world military capability.
🇾🇪 Yemen's Houthis used Claude to build ballistic missiles.
Anthropic says an account linked to Yemen’s Houthi movement used Claude for weapons-related research and development.
The model was reportedly used to help with technical work around ballistic missile systems, alongside other military applications.
Same models that can help write software, solve mathematics and accelerate scientific research can also lower the expertise barrier for building sophisticated weapons.
Frontier models are already becoming tools for real world military capability.
@Kalshi More details here
Claude is getting into dangerous hands.
Anthropic says it shut down accounts linked to biological weapons research, cyberattacks, surveillance, weapons development and influence operations.
In one case, a user sought help with gain-of-function research on chikungunya.
In another, Russian state-linked hackers allegedly used Claude to help develop malware and attack Ukrainian government targets.
Other cases involved missile systems, drones, intelligence gathering and political influence campaigns.
The scary part is that people are increasingly getting enough value from AI to build real operations around it.
@Polymarket Here's what happened
Claude is getting into dangerous hands.
Anthropic says it shut down accounts linked to biological weapons research, cyberattacks, surveillance, weapons development and influence operations.
In one case, a user sought help with gain-of-function research on chikungunya.
In another, Russian state-linked hackers allegedly used Claude to help develop malware and attack Ukrainian government targets.
Other cases involved missile systems, drones, intelligence gathering and political influence campaigns.
The scary part is that people are increasingly getting enough value from AI to build real operations around it.
Claude is getting into dangerous hands.
Anthropic says it shut down accounts linked to biological weapons research, cyberattacks, surveillance, weapons development and influence operations.
In one case, a user sought help with gain-of-function research on chikungunya.
In another, Russian state-linked hackers allegedly used Claude to help develop malware and attack Ukrainian government targets.
Other cases involved missile systems, drones, intelligence gathering and political influence campaigns.
The scary part is that people are increasingly getting enough value from AI to build real operations around it.
@WatcherGuru Here's more
Claude was hiding inside Chinese AI.
Anthropic says Chinese labs including DeepSeek, Moonshot and others used thousands of fraudulent accounts and proxy networks to secretly extract Claude’s capabilities.
Moonshot allegedly generated 3.4M+ exchanges with Claude.
DeepSeek generated 150,000+.
MiniMax generated 13M+.
Reuters reports, Anthropic has identified nearly 190M interactions across seven Chinese labs in the broader campaign.
Anthropic says some operations used proxy networks with 20,000+ fraudulent accounts, making the traffic look like normal users.
So, while users thought they were talking to Kimi or DeepSeek, some of the intelligence powering those systems was allegedly coming from Claude on the backend.
@IntCyberDigest Here's what happened
Claude was hiding inside Chinese AI.
Anthropic says Chinese labs including DeepSeek, Moonshot and others used thousands of fraudulent accounts and proxy networks to secretly extract Claude’s capabilities.
Moonshot allegedly generated 3.4M+ exchanges with Claude.
DeepSeek generated 150,000+.
MiniMax generated 13M+.
Reuters reports, Anthropic has identified nearly 190M interactions across seven Chinese labs in the broader campaign.
Anthropic says some operations used proxy networks with 20,000+ fraudulent accounts, making the traffic look like normal users.
So, while users thought they were talking to Kimi or DeepSeek, some of the intelligence powering those systems was allegedly coming from Claude on the backend.
Most Popular Users
Elon Musk
@elonmusk
241.7M followers
Barack Obama
@barackobama
119M followers
Cristiano Ronaldo
@cristiano
114.2M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.2M followers
Rihanna
@rihanna
98.7M followers
NASA
@nasa
92.4M followers
Justin Bieber
@justinbieber
91.8M followers
KATY PERRY
@katyperry
89.9M followers
Taylor Swift
@taylorswift13
83.8M followers
Lady Gaga
@ladygaga
75.3M followers
Virat Kohli
@imvkohli
73.2M followers
Kim Kardashian
@kimkardashian
70.8M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
66.2M followers
Bill Gates
@billgates
65.1M followers
Selena Gomez
@selenagomez
63M followers
The Ellen Show
@theellenshow
62.3M followers
CNN
@cnn
61.8M followers
X
@x
60.7M followers