Joshua Loftus
@joft.bsky.social
New father. Also: Prof. of #Statistics, #machinelearning / ethics for #datascience @LSE. Unschooled to community college to PhD @Stanford
Technology, institutions, and ideas should serve people, but much of humanity is stuck with this upside-down.
Technology, institutions, and ideas should serve people, but much of humanity is stuck with this upside-down.
trump is still slandering Alex Pretti while the New York Times runs propaganda pieces trying to reconcile him with centrists
February 1, 2026 at 10:26 PM
trump is still slandering Alex Pretti while the New York Times runs propaganda pieces trying to reconcile him with centrists
Lazily googling to find the title of a Poe story instead of going to my bookshelf
Instead of directing me to A Tale of the Ragged Mountains (good), google generates an imitation of a Poe story (🫥)
Instead of directing me to A Tale of the Ragged Mountains (good), google generates an imitation of a Poe story (🫥)
January 7, 2026 at 10:12 AM
Lazily googling to find the title of a Poe story instead of going to my bookshelf
Instead of directing me to A Tale of the Ragged Mountains (good), google generates an imitation of a Poe story (🫥)
Instead of directing me to A Tale of the Ragged Mountains (good), google generates an imitation of a Poe story (🫥)
Tired half the time and have headaches most days
And honestly, don't mind it at all because of this guy's smile
And honestly, don't mind it at all because of this guy's smile
December 30, 2025 at 9:23 PM
Tired half the time and have headaches most days
And honestly, don't mind it at all because of this guy's smile
And honestly, don't mind it at all because of this guy's smile
After a month in Sicily
My stomach just texted me this??? how
My stomach just texted me this??? how
December 7, 2025 at 2:49 PM
After a month in Sicily
My stomach just texted me this??? how
My stomach just texted me this??? how
This is so much more scandalous than some UCs offering remedial math courses
But somehow I won't be surprised if the same journos and pundits who're so concerned about "woke" lowering standards are silent about this
www.nber.org/system/files...
But somehow I won't be surprised if the same journos and pundits who're so concerned about "woke" lowering standards are silent about this
www.nber.org/system/files...
November 24, 2025 at 11:29 AM
This is so much more scandalous than some UCs offering remedial math courses
But somehow I won't be surprised if the same journos and pundits who're so concerned about "woke" lowering standards are silent about this
www.nber.org/system/files...
But somehow I won't be surprised if the same journos and pundits who're so concerned about "woke" lowering standards are silent about this
www.nber.org/system/files...
I'm getting cautiously optimistic that the AI insanity will be ended soon by (1) a deluge of lawsuits and (2) society in general learning that it's bullshit
on.ft.com/3M4Y61L
www.nytimes.com/2025/11/12/b...
on.ft.com/3M4Y61L
www.nytimes.com/2025/11/12/b...
November 23, 2025 at 9:13 PM
I'm getting cautiously optimistic that the AI insanity will be ended soon by (1) a deluge of lawsuits and (2) society in general learning that it's bullshit
on.ft.com/3M4Y61L
www.nytimes.com/2025/11/12/b...
on.ft.com/3M4Y61L
www.nytimes.com/2025/11/12/b...
Logging on to X the Everything app™️ and having the audacity to say remedial math classes are good actually and universities should educate and not just select students
November 22, 2025 at 11:07 AM
Logging on to X the Everything app™️ and having the audacity to say remedial math classes are good actually and universities should educate and not just select students
November 11, 2025 at 1:14 PM
Me, to the universe: you're welcome. It was easy actually
October 23, 2025 at 9:53 PM
Me, to the universe: you're welcome. It was easy actually
Over 300,000 children have died so far as a direct result of trump and musk illegally defunding USAID
September 18, 2025 at 8:40 PM
Over 300,000 children have died so far as a direct result of trump and musk illegally defunding USAID
Left: The entire British political establishment working diligently to destroy Jeremy Corbyn and keep him out of power by any means
Right: Starmer's Labour, in power, manages to become more unpopular than its lowest point under Corbyn
Right: Starmer's Labour, in power, manages to become more unpopular than its lowest point under Corbyn
September 15, 2025 at 6:27 PM
Left: The entire British political establishment working diligently to destroy Jeremy Corbyn and keep him out of power by any means
Right: Starmer's Labour, in power, manages to become more unpopular than its lowest point under Corbyn
Right: Starmer's Labour, in power, manages to become more unpopular than its lowest point under Corbyn
Noahpinion accidentally admitting that the left isn't responsible for cancel culture, but probably doesn't even realize he implied as much
September 13, 2025 at 9:09 PM
Noahpinion accidentally admitting that the left isn't responsible for cancel culture, but probably doesn't even realize he implied as much
This is what I've been saying since 2023 (image below)
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
September 2, 2025 at 8:31 AM
This is what I've been saying since 2023 (image below)
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
"prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"
I'm guessing the general pattern is that these DOGE losers asked "grok" to summarize/search documents for them, didn't check any details, then made life and death decisions based on "hallucinations"
August 22, 2025 at 2:42 PM
I'm guessing the general pattern is that these DOGE losers asked "grok" to summarize/search documents for them, didn't check any details, then made life and death decisions based on "hallucinations"
I know I didn’t hit depth but I’m counting it anyway
July 26, 2025 at 10:06 PM
I know I didn’t hit depth but I’m counting it anyway
Traditional Irish apizza
June 21, 2025 at 3:55 PM
Traditional Irish apizza
Developing a new phobia
May 21, 2025 at 7:24 PM
Developing a new phobia
google AI is like google maps (created >20 years ago) but built on trillions of dollars of intellectual property theft, billions in environmentally costly computation, and also wrong
(the A6 connects Belfast and Derry, the M1 does not)
(the A6 connects Belfast and Derry, the M1 does not)
May 20, 2025 at 8:40 PM
google AI is like google maps (created >20 years ago) but built on trillions of dollars of intellectual property theft, billions in environmentally costly computation, and also wrong
(the A6 connects Belfast and Derry, the M1 does not)
(the A6 connects Belfast and Derry, the M1 does not)
Is this going to work? Adding 5kg some weeks but lowering the reps of intensity day
My goal is to hit at least a single with 185kg at the end
My goal is to hit at least a single with 185kg at the end
May 19, 2025 at 1:25 PM
Is this going to work? Adding 5kg some weeks but lowering the reps of intensity day
My goal is to hit at least a single with 185kg at the end
My goal is to hit at least a single with 185kg at the end
Hilariously, this actually shows the insurance industry is still not willing to underwrite the *baseline* risks of "AI" tools
It will only cover a drop in performance, e.g. from 5% errors to 15% errors. Hence if it stays constant at baseline those 5% errors won't be covered
It will only cover a drop in performance, e.g. from 5% errors to 15% errors. Hence if it stays constant at baseline those 5% errors won't be covered
May 11, 2025 at 10:58 AM
Hilariously, this actually shows the insurance industry is still not willing to underwrite the *baseline* risks of "AI" tools
It will only cover a drop in performance, e.g. from 5% errors to 15% errors. Hence if it stays constant at baseline those 5% errors won't be covered
It will only cover a drop in performance, e.g. from 5% errors to 15% errors. Hence if it stays constant at baseline those 5% errors won't be covered
