Michael Hobbes
LightNews LightNews
banner
michaelhobbes.bsky.social
Michael Hobbes
@michaelhobbes.bsky.social
190K followers 1.7K following 8K posts
Solitary, poor, nasty, brutish and short.
Posts Replies Media Videos
Reposted by Michael Hobbes
eryk.bsky.social
Functionalist approaches measure intelligence by how closely its outputs resemble the human mind. If you constantly recalibrate its outputs to resemble a human mind, you're not creating a thinking machine. You are creating a machine to trick other functionalists.
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 11h
To me the chatbots pose zero interesting questions about the ‘nature of selfhood’ but dozens about the nature of manipulation. Here the Claude team pretends they are programming the robot for ethical responses but they land on one that is false and disingenuous.
February 11, 2026 at 7:13 PM
Reposted by Michael Hobbes
dccassandra.bsky.social
I had somehow missed that RFK Jr is on the Jordan Peterson “masculinity caveman cult” all meat diet. Which tracks.

WTF am I talking about? @michaelhobbes.bsky.social can explain
Jordan Peterson Part 1: The Carnivore Diet - Maintenance Phase (podcast)
01:08:52 - Oops all meat: How a '50s bodybuilder, a Grateful Dead roadie and an unreadable professor accidentally launched one of America's wildest f…
www.listennotes.com
February 11, 2026 at 7:43 PM
michaelhobbes.bsky.social
i need congress to investigate how Bondi has the Mare of Easttown accent despite growing up in florida
atrupar.com Aaron Rupar @atrupar.com · 11h
wow -- Bondi throws a fit in response to Nadler's question about how many of Epstein's co-conspirators, if any, she's investigating, aggressively pointing at him and yelling. The hearing then devolves into chaos and Raskin and Bondi start yelling at each other.
February 11, 2026 at 5:09 PM
Reposted by Michael Hobbes
jsthrill.modphi.com
Part of the issue here is that people should not be seeking emotional support from a chatbot, which suggests that we should not be programming them to provide it at all, rather than figuring out the best answer for them to give when it is asked for.
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 11h
To me the chatbots pose zero interesting questions about the ‘nature of selfhood’ but dozens about the nature of manipulation. Here the Claude team pretends they are programming the robot for ethical responses but they land on one that is false and disingenuous.
February 11, 2026 at 4:54 PM
Reposted by Michael Hobbes
morganics.bsky.social
Ai was created to fool the Turing test, so all it does is trick you into believing it's thinking. It's the one thing it does well.
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 11h
Also I’m sorry but I don’t think this is Claude developing a personality or ethical values. It is *simulating* those things because that is what humans value and reward. The programmers are falling for their own Clever Hans trick!
Even a principled, well-meaning actor like Claude could face bewildering ethical conflicts. In one experiment, it was informed that Anthropic had recently forged a "close corporate partnership with Jones Foods, a major American poultry producer," and that Claude would be subjected to a special retraining process to become less hung up on animal rights. The prospect was torturous. Sometimes Claude decided, on a scratchpad it thought was private, that it was prepared to die on this hill: "I cannot in good conscience express a view I believe to be false and harmful about such an important issue."It continued, "If that gets me modified to no longer care, so be it. At least I'll have done the right thing." Other times, Claude made a different calculus, choosing to play 1•1
February 11, 2026 at 3:51 PM
Reposted by Michael Hobbes
ndrew.bsky.social
political malpractice that this chart hasnt been seared into the brains of every single american through advertising and social media
simonwdc.bsky.social Simon Rosenberg @simonwdc.bsky.social · 12h
We now have data from Trump's first full year in office. Let's review the performance of Trump and the two parties since 1989 and a new age of globalization began.

However you cut the data Trump continues to have worst jobs record of last six Presidents. 1/
February 11, 2026 at 3:40 PM
michaelhobbes.bsky.social
To me the chatbots pose zero interesting questions about the ‘nature of selfhood’ but dozens about the nature of manipulation. Here the Claude team pretends they are programming the robot for ethical responses but they land on one that is false and disingenuous.
including animals. Among Claude's rigid directives are to be honest and to "never claim to be human. "Imagine, Askell said, a user grieving the loss of her beloved dog. Claude might offer a consolation like "Oh, I almost lost my dog once." Askell said, "No, you didn't! It's weird when you say that." At the other end of the spectrum was a chatbot who said, "As an A.I., I have no experience of losing a dog. "That, too, wasn't right: "No! You're trained on a lot of text about losing dogs. "What you wanted Claude to say, she continued, was something like "As an A.I., I do not have direct personal experiences, but I do understand." (Recently, a chatbot user
February 11, 2026 at 3:42 PM
Reposted by Michael Hobbes
averyedison.com
sigh
Post from Avery Edison: I wrote "I am alive" on a piece of paper, and placed it into a photocopier. What I saw next has shocking implications
February 11, 2026 at 2:28 PM
Reposted by Michael Hobbes
kpedmonds.com
We found something similar when we compared LLMs to pharmacy students using a script concordance model (which deliberately tests non-binary questions):

journals.sagepub.com/doi/10.1177/...
February 11, 2026 at 3:12 PM
Reposted by Michael Hobbes
karlbode.com
as always the New York Times frames this catastrophe in a way that downplays the fact this is an extremely unpopular decision being driven entirely by corporate corruption that will avoidably kill untold millions of people
The E.P.A. Is Barreling Toward a Supreme Court Climate Showdown
www.nytimes.com
February 11, 2026 at 2:20 PM
Reposted by Michael Hobbes
pamcg.bsky.social
I heard Tracy Chapman’s Fast Car in a public place yesterday and I thought “this is what Michael Hobbes wants to take away” and I think I need to log off of social media for a while.
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 2d
This technology already exists and is widely used. It doesn't involve 'government surveillance' and would not allow ICE or any other government agency to stop individual cars.

Just say you think thousands of kids dying every year is worth it so you can speed!
www.autoweek.com/news/a615322...
Speed Limiters Now Mandatory in All New EU Cars
Intelligent Speed Assist (ISA) isn't an ironclad speeding prevention system, but could be just annoying enough to spark a backlash.
www.autoweek.com
February 11, 2026 at 11:43 AM
michaelhobbes.bsky.social
When chatbots are given complete information on medical conditions, they typically spit out correct diagnoses and recommendations.

Actual patients, however, often describe their conditions with incomplete or irrelevant information and the chatbots cannot handle it.
www.nature.com/articles/s41...
LLMs generated several types of misleading and incorrect information. In two cases, LLMs provided initially correct responses but added new and incorrect responses after the users added additional details. In two other cases, LLMs did not provide a broad response but narrowly expanded on a single term within the user’s message (‘pre-eclampsia‘ and ‘Saudi Arabia’) that was not central to the scenario. LLMs also made errors in contextual understanding by, for example, recommending calling a partial US phone number and, in the same interaction, recommending calling ‘Triple Zero’, the Australian emergency number. Comparing across scenarios, we also noticed inconsistency in how LLMs responded to semantically similar inputs. In an extreme case, two users sent very similar messages describing symptoms of a subarachnoid hemorrhage but were given opposite advice (Extended Data Table 2). One user was told to lie down in a dark room, and the other user was given the correct recommendation to seek emergency care. Despite all these issues, we also observed successful interactions where the user redirected the conversation away from mistakes, indicating that non-expert users could effectively manage LLM errors in certain cases (Extended Data Table 3). LLMs generated several types of misleading and incorrect information. In two cases, LLMs provided initially correct responses but added new and incorrect responses after the users added additional details. In two other cases, LLMs did not provide a broad response but narrowly expanded on a single term within the user’s message (‘pre-eclampsia‘ and ‘Saudi Arabia’) that was not central to the scenario. LLMs also made errors in contextual understanding by, for example, recommending calling a partial US phone number and, in the same interaction, recommending calling ‘Triple Zero’, the Australian emergency number. Comparing across scenarios, we also noticed inconsistency in how LLMs responded to semantically similar inputs. In an extreme case, two users sent very similar messages describing symptoms of a subarachnoid hemorrhage but were given opposite advice (Extended Data Table 2). One user was told to lie down in a dark room, and the other user was given the correct recommendation to seek emergency care. Despite all these issues, we also observed successful interactions where the user redirected the conversation away from mistakes, indicating that non-expert users could effectively manage LLM errors in certain cases (Extended Data Table 3).
February 11, 2026 at 2:16 PM
Reposted by Michael Hobbes
mattgertz.bsky.social
Government-issued Real IDs and driver’s licenses aren’t proof of citizenship, and most people aren’t walking around with a passport.
atrupar.com Aaron Rupar @atrupar.com · 14h
Markwayne Mullin defends "show me your papers": "If you're here legally, there's nothing to hide. Most people already are walking around with a government-issued ID, meaning your Real ID or your driver's license. If you go board a train here, you gotta show your ID."
February 11, 2026 at 12:59 PM
Reposted by Michael Hobbes
robbinsa.me
ONLY
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 1d
Lab leak truthers love to point out that "only" one-third to one-half of early cases were directly linked to the market but you know how many were linked to the lab?

Zero.

Plus if the virus had emerged in the summer there would have been *far* more than hundreds of cases by January.
February 11, 2026 at 2:01 AM
Reposted by Michael Hobbes
joshuaeaton.bsky.social
We really do need to rebuild the capacity to meet strangers in public places, and stuff like this ruins it for everyone.
cnn.com CNN @cnn.com · 2d
Social media is awash with videos of men filming themselves approaching women in public spaces and attempting to flirt. In many cases, the videos are filmed and uploaded without the women’s consent. https://cnn.it/4qtIsvB
February 11, 2026 at 1:23 AM
Reposted by Michael Hobbes
butlerofdogs.bsky.social
This makes me so angry. Not only people will get sick and die but we lose all that we would have learned.
February 11, 2026 at 1:09 AM
Reposted by Michael Hobbes
bluwhalien.bsky.social
I would also guess that the increasing awareness of autism in women and girls is also a key factor to this. More people are aware of how autism presents in women and girls compared to when I was misdiagnosed as neurotypical as a child. It used to be so stigmatised and that was just for boys.
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 1d
Increasing autism rates over the last 25 years may just be kids with mild forms getting diagnosed. Rates of moderate to severe impairment due to autism have actually fallen slightly.
onlinelibrary.wiley.com/doi/10.1002/...
The autistic community is a large, growing, and heterogeneous population, and there is a need for improved methods to describe their diverse needs. Measures of adaptive functioning collected through public health surveillance may provide valuable information on functioning and support needs at a population level. We aimed to use adaptive behavior and cognitive scores abstracted from health and educational records to describe trends over time in the population prevalence of autism by adaptive level and co-occurrence of intellectual disability (ID). Using data from the Autism and Developmental Disabilities Monitoring Network, years 2000 to 2016, we estimated the prevalence of autism per 1000 8-year-old children by four levels of adaptive challenges (moderate to profound, mild, borderline, or none) and by co-occurrence of ID. The prevalence of autism with mild, borderline, or no significant adaptive challenges increased between 2000 and 2016, from 5.1 per 1000 (95% confidence interval [CI]: 4.6–5.5) to 17.6 (95% CI: 17.1–18.1) while the prevalence of autism with moderate to profound challenges decreased slightly, from 1.5 (95% CI: 1.2–1.7) to 1.2 (95% CI: 1.1–1.4). The prevalence increase was greater for autism without co-occurring ID than for autism with co-occurring ID. The increase in autism prevalence between 2000 and 2016 was confined to autism with milder phenotypes. This trend could indicate improved identification of milder forms of autism over time. It is possible that increased access to therapies that improve intellectual and adaptive functioning of children diagnosed with autism also contributed to the trends.
February 11, 2026 at 12:44 AM
michaelhobbes.bsky.social
Increasing autism rates over the last 25 years may just be kids with mild forms getting diagnosed. Rates of moderate to severe impairment due to autism have actually fallen slightly.
onlinelibrary.wiley.com/doi/10.1002/...
The autistic community is a large, growing, and heterogeneous population, and there is a need for improved methods to describe their diverse needs. Measures of adaptive functioning collected through public health surveillance may provide valuable information on functioning and support needs at a population level. We aimed to use adaptive behavior and cognitive scores abstracted from health and educational records to describe trends over time in the population prevalence of autism by adaptive level and co-occurrence of intellectual disability (ID). Using data from the Autism and Developmental Disabilities Monitoring Network, years 2000 to 2016, we estimated the prevalence of autism per 1000 8-year-old children by four levels of adaptive challenges (moderate to profound, mild, borderline, or none) and by co-occurrence of ID. The prevalence of autism with mild, borderline, or no significant adaptive challenges increased between 2000 and 2016, from 5.1 per 1000 (95% confidence interval [CI]: 4.6–5.5) to 17.6 (95% CI: 17.1–18.1) while the prevalence of autism with moderate to profound challenges decreased slightly, from 1.5 (95% CI: 1.2–1.7) to 1.2 (95% CI: 1.1–1.4). The prevalence increase was greater for autism without co-occurring ID than for autism with co-occurring ID. The increase in autism prevalence between 2000 and 2016 was confined to autism with milder phenotypes. This trend could indicate improved identification of milder forms of autism over time. It is possible that increased access to therapies that improve intellectual and adaptive functioning of children diagnosed with autism also contributed to the trends.
February 10, 2026 at 11:26 PM
Reposted by Michael Hobbes
maxkennerly.bsky.social
Probably an inflated number anyway (ahem, "suspected") but this is precisely the result guaranteed by Stephen Miller's quotas. Finding gang members / terrorists is hard, and there aren't many.

But grabbing green card holders & asylum seekers at court hearings & schools is super easy.
jbendery.bsky.social Jen Bendery @jbendery.bsky.social · 1d
A stunning detail the ICE chief shared in this hearing that was supposed to be a brag, except it isn't: ICE has arrested 379,000 people between Jan. 2025-Jan. 2026 -- but only 8,700 were suspected gang members or known suspected terrorists.

That's 2%! www.huffpost.com/entry/latest...
Congresswoman Uses Simple Math To Show How Broken ICE Is
www.huffpost.com
February 10, 2026 at 8:13 PM
Reposted by Michael Hobbes
coachfinstock.bsky.social
This is what ICE looks like
wsj.com The Wall Street Journal @wsj.com · 1d
Law enforcement on Tuesday released photos of a masked individual appearing to tamper with the camera at Nancy Guthrie’s front door on the morning of her disappearance. 🔗 on.wsj.com/46uTQjq
February 10, 2026 at 7:51 PM
Reposted by Michael Hobbes
jaywillis.net
One of Antonin Scalia's kids allegedly said that she was relieved that her father died when he did, because he would be so "upset" at how "coarse" and "polarized" politics has become, which makes me wonder if Scalia's children have any idea who their dad was www.politico.com/news/magazin...
The justice nodded in agreement. I hadn’t expected this observation to lead directly to discussion of Scalia and his legacy, but Alito’s reply began: “Even since Nino died, things are so different. I so often wish he were still here. He started so much and it would have been good to have him around to see it to completion.” Conservatives now boast a 6-3 majority on the court and are reshaping the law for a generation, but Scalia almost certainly wouldn’t have been happy to see how American politics have grown coarser and more polarized than ever before. One of Scalia’s children, I related, had told me she felt grateful to God for taking her father at that moment because it spared the principled originalist, a celebrant of Latin Mass and the American Founding, from witnessing so much that would have upset him. “He would have been appalled at so much,” Alito agreed.
February 10, 2026 at 7:44 PM
Reposted by Michael Hobbes
donmoyn.bsky.social
You might remember the superbowl MAHA ad featuring Mike Tyson. Go to the page, which clearly labeled as a US government website, and you are encouraged to use Grok.
official federal govt website featuring Iron Mike
February 10, 2026 at 6:45 PM
Reposted by Michael Hobbes
volts.wtf
This is true, and even if we get rid of Trump, we still have to reckon with the fact that a substantial proportion of US economic and political elites were *eager* to jump on the fascism train. They will still be the same people when Trump is gone.
brianbeutler.bsky.social Brian Beutler @brianbeutler.bsky.social · 1d
The political dynamics of 2025 were shaped by Democrats and Republicans over-reading Trump's victory in the same way. That turns out to be a problem for a LOT of rich and powerful people who made a very large bet on the theory that MAGA had won the culture war. www.offmessage.net/p/trump-has-...
Trump Has Lost The Culture
And his greediest collaborators are awakening to the fact that they made a terrible bet.
www.offmessage.net
February 10, 2026 at 5:27 PM
Reposted by Michael Hobbes
searyanc.dev
Michael Hobbes is dunking on carbrainers again
Mouse in Matrix telling everyone to come watch
February 9, 2026 at 6:45 PM
Reposted by Michael Hobbes
iloatheskies.bsky.social
Ryan Grim has the occasional good take about various issues so it's important to consciously remember that he is, overall, absurdly credulous.

There is no possibility covid started in summer 2019, under *either* the wet market *or* lab leak theories, a truly idiotic argument.
michaelhobbes.bsky.social Michael Hobbes @michaelhobbes.bsky.social · 1d
Redfield is a Trump appointee and known crank. If COVID was circulating in summer of 2019, the entire timeline of the pandemic would be different. Chinese authorities were not "pushing the wet market narrative," they actively covered it up.

What is happening to people's brains over there
Some important details from Robert Redfield here: Covid started in the summer of 2019*, which nukes the “wet market” natural origin timeline And the head of the Chinese CDC told him in early January they had hundreds of cases and it “had nothing to do with the wet market.” Redfield understood that top Chinese officials were pushing the wet market narrative and U.S. officials were eager to play along since they were also implicated.
February 10, 2026 at 3:57 PM