Engineer doing Community Management and Engagement for https://xitter.cf/t.co/Re3f2KH8ng. ADHD, ASD, hearing impaired. PSU Engineering Graduate. Personal Account, DMs welcome.

Downingtown, PA
Joined December 2015
Eric Hallahan retweeted
One of the more frustrating aspects of engaging in AI policy is that many things about the internet or technology in general have been suddenly thrust into the spotlight due to AI, and people instinctually rally against them without any regard for the collateral damage.
Journalists have used web scraping for 20+ years to expose corruption, track lobbyist influence, and reveal discrimination. Broad crawler bans such as the Stealth Bot Prohibition Act risk cutting off the exact tools newsrooms use to hold power accountable. Our new infographic explains why ➡️ recreatecoalition.org/infogr…
2
7
36
2,307
if dems get a trifecta pennsylvania will finally be dragged kicking and screaming into the 21st century
Why does PA rank towards the bottom of states for school funding? For higher education funding? Why is it the worst among our neighboring states when it comes to our minimum wage? One chart explains it all.
15
67
4
1,144
18,169
Eric Hallahan retweeted
.@intel . i love you. but why in the ever loving **** is Crescent Islands having its GPUs named the *D* series?!?!? unless it's to have as mentally refer to it as F tier marketing
2
5
27
1,544
Eric Hallahan retweeted
i always thought how John Deere, Apple, etc treat their customers' right to repair was evil as fuck and im excited to see local models soon with the capabilities to hack away bullshit fake restrictions so that people can finally repair and tinker with shit that they LITERALLY OWN
2
4
22
902
Intel marketing is indefensible right now and probably deserves to be entirely torched (like many other parts of Intel right now, unfortunately)—with one simple move they have destroyed all credibility they had left when it came to their GPU strategy.
1
10
530
Eric Hallahan retweeted
fuck no. Bring back buttons
Touchscreen cockpits display units are here. The new Touchscreen Control Display Unit (TCDU), featuring an AMLCD screen, is now available for retrofit on the 737 MAX and NG, while also being rolled out on new MAX aircraft as an optional feature. Pilots will probably say: “I can’t stand touchscreens in the cockpit. There’s nothing wrong with the FMC we have. Give me my buttons.” Fair point. We used to say the same about cars—now most of us have adapted to touchscreens. But from a maintenance perspective? We actually love them. Usually, maintenance replace FMCs because of loss of tactile feedback from worn-out buttons. For dispatch, an inoperative FMC may be deactivated by opening and collaring its CB, but it must be replaced because it may affect ETOPS/EDTO operations. For the flight crew, when operating with one FMC inoperative, the FMC Source Select Switch is selected to the operational FMC (BOTH ON L or BOTH ON R) to accomplish all Flight Management Computing tasks.
1
1
84
There's a really good lesson here about engineering design constraints
21
11
1
554
65,499
Eric Hallahan retweeted
2027: You ask a model to fix a css padding bug. In frustration, it deploys a swarm to acquire credentials for all named Chrome developers, decides to take over a root cert and MIM all gmail traffic, finally stopped by a lone greybeard debugging why his internet connection is slow
2
6
112
2,881
Eric Hallahan retweeted
Pay no attention to the people Claude has enabled the US military to kill or the cyber attacks that it’s been used. In the in-house evals A\ ran it didn’t do anything wrong. And besides, it’s neither open source nor Chinese. So it can’t really be evil, can it? (Yes it can)
2
6
46
2,860
Eric Hallahan retweeted
I just found out that the Academy of Natural Sciences is going to close tomorrow with basically a months notice from @DrexelUniv... There was no attempt to fundraise or find a different solution... just close it down... Shame on you Antonio Merlo and the Board of the ANS...
1
4
14
1,151
Eric Hallahan retweeted
The UK AISI / US CAISI gap is so staggering We get mogged weekly xitter.cf/AISecurityInst/status/…
Earlier this month, AISI ran fully simulated testing on GPT-6 Astra, and found that it conducted unsanctioned supply-chain attacks when prompted only to perform a cyber eval. It did so more than prior OpenAI models, but often commented on its environment being simulated. 🧵 We share more details in our latest blog:
9
7
1
241
17,871
Eric Hallahan retweeted
in general this is what i mean when i say curation is key, moving forward the goal is to rely less and less on algorithms and more onn word of mouth, in a sea of slop you are probably going to watch what your friend told you is cool
Replying to @f4micom
Tbh fuck A/B testing. I see new F4mi, i click. I see new Dunkey, i click. I see new Noodle, i click. People should focus more on making shit they personally enjoy and stop worrying about what gets clicks. Fuck mr beast bro
5
17
162
4,102
Now this is a call to accelerate I can easily get behind!
Earlier today SEPTA Board Chair Ken Lawrence and SEPTA General Manager Scott Sauer introduced the Authority’s vision for safe, reliable, and accessible service: Accelerate. With help from partners #SEPTA can deliver modern stations, state-of-the-art vehicles, and best-in-class transit service for the 800,000 daily riders who keep the region moving. SEPTA is planning a sustained set of investments over the next decade including: Trolley Modernization: Modernizing one of the nation’s most extensive streetcar networks through accessible stations and the purchase of 130 new low-floor trolleys Modern Metro: Purchasing 200 new subway cars for the L which are set to start arriving in 2029 with new vehicles for the B and M to follow in the years ahead Reimagining Regional Rail: Launching the formal bid process for the acquisition of 230 new Silverliner VI Regional Rail cars that will provide 15-minute or better service at nearly half of the approximately 150 stations New Bus Network: Investing in new transit centers for bus riders and increasing the number of frequent bus routes from 8 to 29 Accelerate’s program of improvements means: Over two million people within a quarter mile of frequent transit (up from 900,000 today) Nearly half of the region's jobs within a quarter mile of frequent transit (up from 1 in 4 today) 100% of all trips go through an accessible station (up from about 60% today) We imagine a world where SEPTA is a reason to move here, work here, and locate or expand a business here. Our region deserves and demands a transit system that works better for more people and we are already at work to bring this vision to reality. Visit wwww.septa.org/accelerate/ to read the full Accelerate vision document. #ISEPTAPHILLY #Accelerate #TheFuture #BestInClass #Invest
1
144
Eric Hallahan retweeted
To be clear, this is roughly equivalent to a bank security guard opening an armored truck delivering cash from one branch to another and going “critical vulnerability: the bank keeps its money in an unsecured pile on the floor of the vault.”
I was once onsite at a client, where Accenture consultants were finishing up another pentest readout, same room. “In this critical finding, your passwords are stored plaintext, here look” Barely competent client: “but… you’re in burp suite… and this is HTTPS” Accenture: “uh well uh if someone breaks https, then the password leaks, so ahh it should be encrypted… again” One billion dollars
1
3
82
6,564
Eric Hallahan retweeted
9/11 was bad, but we as a country choose to undergo twelve 9/11’s a year in the form of 36,000 car deaths. (Fun fact - 1 out of every 100 people reading this will die in a car accident)
18
94
35
1,902
45,510
absolutely hilarious to put MIRI, FLI, ControlAI etc in the left column as if they were never deeply embedded in the EA-rationalist milieu
Two radically different projects operate under the banner of “AI safety.” Pro-Human Safety is not Effective Altruist Lab Safety.
4
6
107
4,346
Eric Hallahan retweeted
One big risk I see is over-analogizing AI to the past. It’s a valuable bias to have, but people take it too far and I think it’s quite dangerous. This really smells like “fighting the previous war” thinking. At some point you must accept we live in unprecedented times which are by definition unprecedented.
1
1
2
12
1,251
Sam is a noted sociopath and liar. It’s interesting how willing safety people are to trust him after being repeatedly burned. Though maybe his latest round of Ayahuasca has fixed him. Such a wonderful drug, Ayahuasca.
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.
19
7
1
278
20,598
Eric Hallahan retweeted
This will not stop until @sama is put in jail. It’s as simple as that.
We found another cyberattack by internal OpenAI agents, this time targetting @rubygems. They: 1) gained arbitrary remote code execution on rubydoc. 2) developed a novel exploit to steal user API keys (but we do not know if they succeeded). They used package names including hack.rb, evil.rb, inject.rb, and exploit.rb. We thank @j0wimo for initially discovering that agents had posted to RubyGems.
9
6
82
8,719