Figure 5 is really quite remarkable. It seems to show that normal LLMs are better at tasks where the correct answer is likely to be the next token. For tasks that require a small number of intermediate steps, current reasoning models do much better, but break down as the number of intermediate steps grow.
This seems to indicate that the next generation of models should focus on recursively solving small parts of the problem before function-calling another model to solve another small part of the problem and working it's answer into the reasoning loop.
Many seem to be citing this paper as an indication that LLMs are over - I think this indicates a clear path towards the next step function change in their abilities.
Respondology is working on a social media comment activation platform, building tools to automatically hide spam, abuse, and hate comments, understand audiences, and respond to users using GenAI/LLMs. We are a small team that just raised a recent funding round and are rapidly growing. We are looking to hire a Senior Backend Engineer with significant experience with Python, experience with Opensearch/Elasticsearch is also a plus [120k-140k + equity + bonus]. Our core tech stack is: React, a core Ruby-on-Rails monolith with Python+FastAPI microservices, Postgres, AWS. We are based out of Boulder, CO, but the engineering team is distributed across the US. Learn more here: https://respondology.com/careers/
The last Starship test flight failed - now SpaceX is probably not going to make the Mars launch window next year that they were aiming to hit. It will also probably not meet the goal NASA had to return to the moon in 2027.
The AP article [1] has the full complaint linked, the crux of the case seems to be around the judge allowing the defendant to leave through a back entrance ("jury door") when they were aware agents were waiting in the public hallway to make an arrest as they exited.
" 29. Multiple witnesses have described their observations after Judge DUGAN returned to her courtroom after directing members of the arrest team to the Chief Judge’s office. For example, the courtroom deputy recalled that upon the courtroom deputy’s return to the courtroom,defense counsel for Flores-Ruiz was talking to the clerk, and Flores-Ruiz was seated in the jury box, rather than in the gallery. The courtroom deputy believed that counsel and the clerk were having an off-the-record conversation to pick the next court date. Defense counsel and Flores-Ruizthen walked toward each other and toward the public courtroom exit. The courtroom deputy then saw Judge DUGAN get up and heard Judge DUGAN say something like “Wait, come with me.” Despite having been advised of the administrative warrant for the arrest of Flores-Ruiz, Judge DUGAN then escorted Flores-Ruiz and his counsel out of the courtroom through the “jury door,” which leads to a nonpublic area of the courthouse. These events were also unusual for two reasons.First, the courtroom deputy had previously heard Judge DUGAN direct people not to sit in the jury box because it was exclusively for the jury’s use. Second, according to the courtroom deputy, only deputies, juries, court staff, and in-custody defendants being escorted by deputies used the back jury door. Defense attorneys and defendants who were not in custody never used the jury door."
I find it extremely doubtful that she told the agents he'd be leaving through a particular door or that she had any legal obligation to make sure the man exited in a particular way.
United States Code, Title 8, § 1324(a)(1)(A)(iii) (2023)
> (1)(A) Any person who
[…]
> (iii) knowing or in reckless disregard of the fact that an alien has come to, entered, or remains in the United States in violation of law, conceals, harbors, or shields from detection, or attempts to conceal, harbor, or shield from detection, such alien in any place, including any building or any means of transportation;
[…]
> shall be punished as provided in subparagraph (B).
>attempts to conceal, harbor, or shield from detection, such alien
Does that cover "Hey, this door is closer to where we are going"? It's going to rest on convincing a jury that the only possible reason the suspect would go out that door would be the judge explicitly trying to help them evade arrest.
IMO that should be impossible to prove but we have never taken "Beyond a reasonable doubt" seriously.
"knowing or in reckless disregard"
Funny, nobody ever arrests any of the employers choosing not to verify the documents of their employees.
There was a series of other steps in the case before she let the guy use a door the public never uses. Including rushing his case through after confronting the agents and adjourning it without notifying the attorneys present in the court room. She instructed him through the door before the basic administrative tasks of the case were done.
If the details of what their witness (court deputy) says is true then it's pretty obvious what happened here.
Whether the government should give State judges that leeway is another issue.
We really need a new HN. The fact that people will downvote your comment, which is literally 100% correct, simply because they want to believe a falsehood, is pretty sad. This isn't Hacker News, it's Delusion News.
Another possibility is for downvotes to be public, and then allow people to "filter out" downvoters who seem politically motivated (a personal list would be fine, I assume it's like 5 people). HN can't really survive in the year 2025 without some radical anti-politics technology.
The usual HN solution to this problem is to just flag purely political stories and not discuss politics.
I agree any system that ranks posts purely on the ratio of upvotes to downvotes isn't well equipped to handle discussions of controversial topics. Unless there's a strong culture of respecting dissenting opinions it inevitably just turns into an echo chamber for one side or the other. There needs to be some way to filter out votes motivated by ideology rather than post quality.
X's Community Notes has the right idea I think, in that its algorithm takes into account the ideological biases of the voters and ranks notes based on the overall consensus across multiple ideological perspectives rather than just on whichever ideological perspective has the greatest total number of votes. That's a lot harder to implement though.
> Milwaukee County Circuit Court Judge Hannah Dugan is accused of escorting the man and his lawyer out of her courtroom through the jury door last week after learning that immigration authorities were seeking his arrest.
Perhaps you’re a lawyer with greater insight into this issue than I have. The actions described in the article satisfy the plain reading of the terms “conceal, harbor, or shield from detection.”
The intuitive reading seems to be further corroborated by the case law:
> The word "harbor" […] means to lodge or to aid or to care for one who is secreting himself from the processes of the law. The word "conceal" […] means to hide or to secrete or to keep out of sight or to aid in preventing the discovery of one who is secreting himself from the processes of the law.
> *The statute proscribes acts calculated to obstruct the efforts of the authorities to effect arrest of the fugitive,* but it does not impose a duty on one who may be aware of the whereabouts of the fugitive, although having played no part in his flight, to reveal this information on pain of criminal prosecution.
The alien was already "detected", that's why they were at the courthouse. She didn't harbor him, she was performing her job and the defendant was required to be there.
I also fail to see how she "concealed" him either. "Aiding and abetting" would be a stretch, but still more accurate verbs. But those aren't in the law you quoted.
Original title: "Former New Mexico judge and wife arrested by ICE". It's as though he wasn't an active judge while harboring an alleged Tren De Aragua gang member.
Protip: believe absolutely nothing you read in mainstream news sources on any even remotely political topic. Read between the lines, sort of like people used to read Pravda in the Soviet Union.
This is why I don't believe 90% of what they say about Trump. As soon as he took office in 2017 it was so obvious that MSM was paid to absolutely destroy him. Then in 2021 suddenly everyone single problem wasn't due to the president anymore.
I woke up to this in 2016 when theretofore beloved public figure Donald Trump turned into literally Hitler immediately after he descended down that elevator in the Trump tower, all without changing a single opinion he'd ever held. And it's been unrelenting ever since.
and he was a former democrat! I'm not saying he's perfect, I'm not saying he's sophisticated, or that he hasn't said some stupid things. No way, there are way better republican candidates. But TDS is real.
What's being alleged is that she deliberately escorted the man out through an exit that is not usually made available to members of the public, instead of allowing him to leave through the regular door that would likely have put him right into the hands of ICE. If that was done with the intent of helping him evade arrest (which, if the story above is accurate, seems likely), it seems very reasonable to charge her with obstruction.
None of that is to say that what she did was morally wrong—often the law and morality are at odds.
You bring up an important point - laws and morality are not equivalent, and often differ. I wish people on HN would get that through their thick skulls and stop downvoting people for saying that a law was broken, if they think the law is invalid. The downvote button is not the ballot box where you vote on which laws you support.
An administrative warrant is still valid for arrest, it doesn't need to be signed by a judge. If you think it needs to be then laws needs to be changed, but that is how laws are right now.
It's not valid for arrest on private premises, like a courthouse. Only out in public.
So the judge is no more "obstructing justice" than if I refuse to open the door to let ICE agents in to arrest someone with an administrative warrant, and then that person leaves out the back door.
ICE should have obtained a judicial warrant, but they didn't. That's their fault.
From reading the affidavit it’s clear to me there is a lot of uncertainty and confusion around these situations. Clearly the judges are upset with ICE making arrests in the public spaces within the court hall while ICE views it as the perfect place since the defendant will be unarmed. This was an administrative warrant and IANAL but doesn’t that not require local cooperation e.g the judge is in her right to not help or comply with the warrant?
If it can be proven that she deliberately escorted the person through the non-public exit to the courtroom with the intent of helping them evade arrest by officers with a warrant who were waiting outside at the other entrance, how would that not be an arrestable offence?
We're extremely light on facts right now, so I'm not taking the above quoted story at face value, but if one were to take it at face value it seems pretty clear cut.
The part that makes it not so clear cut is that this is really a constitutional issue rather than a criminal one. It is not credible that the judge was trying to ensure the man could keep living in the US undocumented. She was defending her court. From a judge's POV, arresting a plantiff/defendant in the middle of a trial is a violation of their right to a trial and impedes local prosecutors' abilities to seek justice.
Ultimately this seems like Trump asserting that the federal executive branch has unfettered veto authority over local judicial branches. That doesn't sit well with me.
The charges against her are not about her behavior during her court, they're about what happened once the court was adjourned and the defendant was starting to leave. She successfully defended the process within her own courtroom and it's alleged that she went a step further in securing the defendant from their impending arrest on an unrelated charge.
There's definitely a conversation to be had about whether people should be safe from immigration law enforcement while within a courthouse, but at the moment as I understand it that is not a protection that exists.
It doesn't matter if court was adjourned, she was still at work and performing her official duties. In particular she asked the ICE agents if they had a judicial warrant and was told no, it was administrative. A federal judicial warrant clearly outranks a local judge and it would be fine to arrest her if she defied it. It is not clear that an order from the executive branch does the same. That doesn't mean local judges are immune from federal prosecution (e.g. corruption charges if someone takes money to rule favorably), but there is a fairly high bar, I don't think her behavior even comes close to probable cause for obstruction of justice or shielding an undocumented immigrant.
The reason every other administration besides Trump refused to go into local courthouses to deport people wasn't about "whether people should be safe from immigration enforcement," it was about separation of powers.
I guess I would hope it takes more then that to rise to the level of obstruction.
Additionally, you have to prove intent, and unless the judge was careless, I doubt they will ever be able to do that. I'd bet the prosecution knows this, I don't think they expect the charges to stick. It seems this arrest was done to send a message.
More evidence of that is that typically in this kind of case they would invite her to show up somewhere to accept process, not be arrested like a common criminal.
In the examples they demonstrate tool use in the reasoning loop. The models pretty impressively recognize they need some external data, and either complete a web search, or write and execute python to solve intermediate steps.
To the extent that reasoning is noisy and models can go astray during it, this helps inject truth back into the reasoning loop.
Is there some well known equivalent to Moores Law for token use? We're headed in a direction where LLM control loops can run 24/7 generating tokens to reason about live sensor data, and calling tools to act on it.
SoftBank was founded by Masayoshi Son. It was originally a tech company that resold software (imagine Steam but for desktop apps), but very quickly it morphed into something like a PE firm that would do continual leveraged buyouts of existing firms. They've bought companies that run conferences, tech magazines, one of their more formative M & A deals was to buy the Japanese Vodafone arm. Many of these deals actually went pretty poorly, with Masayoshi overpaying for the firms he bought and then being saddled with very large debts to repay the loans. Masa was wiped out and faced bankruptcy multiple times.
Later on the would move away from leveraged buyouts and start investing in startups more similar to how a VC firm would in the US. He notably invested early in Yahoo and Alibaba.
The company is not a sovereign fund - it's really run at the behest of it's founder. If you want to read more about the story, there was recently a great biography that chronicled the story: Gambling Man (https://www.amazon.com/Gambling-Man-Greatest-Disruptor-Masay...)
> old power in Japan) has been spending the money on recently
Investing in SoftBank.
MUFG, the Custody Bank, and Master Trust all have significant ownership stakes in SoftBank, and most Japanese conglomerates like NTT, NEC, etc tend to work closely with SoftBank on strategic deals
Yes, Ethnically Korean, his grandparents were immigrants. As for cultural differences, I think you could credit his extensive time spent living in the US as much as the immigrant Korean culture.
Japan has a weird split between Japanese nationals, Korean special residents, and actual foreigners (including "normal" Korean immigrants).
It's a whole can of worm I'll only leave here for any else to open, but Son Masayoshi was born in Japan from parents who might have formerly been Japanese (Korea was a Japanese colony) depending on the circumstances.
Many people in similar cases are basically undistinguishable from regular Japanese nationals culturally speaking.
I'm not an ML researcher, but I do work in the field.
My mental model of AI advancements is that of a step function with s-curves in each step [1]. Each time there is an algorithmic advancement, people quickly rush to apply it to both existing and new problems, demonstrating quick advancements. Then we tend to hit a kind of plateau for a number of years until the next algorithmic solution is found. Examples of steps include, AlexNet demonstrating superior image labeling, LeCun demonstrating DeepLearning, and now OpenAI demonstrating large transformer models.
I think in the past, at each stage, people tend to think that the recent progress is a linear or exponential process that will continue forward. This lead to people thinking self driving cars were right around the corner after the introduction of DL in the 2010s, and super-intelligence is right around the corner now. I think at each stage, the cusp of the S-curve comes as we find where the model is good enough to be deployed, and where it isn't. Then companies tend to enter a holding pattern for a number of years getting diminishing returns from small improvements on their models, until the next algorithmic breakthrough is made.
Right now I would guess that we are around 0.9 on the S curve, we can still improve the LLMs (as DeepSeek has shown wide MoE and o1/o3 have shown CoT), and it will take a few years for the best uses to be brought to market and popularized. As you mentioned, LeCun points out that LLMs have a hallucination problem built into their architecture, others have pointed out that LLMs have had shockingly few revelations and breakthroughs for something that has ingested more knowledge than any living human. I think future work on LLMs are likely to make some improvement on these things, but not much.
I don't know what it will be, but a new algorithm will be needed to induce the next step on the curve of AI advancement.
> Each time there is an algorithmic advancement, people quickly rush to apply it to both existing and new problems, demonstrating quick advancements. Then we tend to hit a kind of plateau for a number of years until the next algorithmic solution is found.
That seems to be how science works as a whole. Long periods of little progress between productive paradigm shifts.
It's been described as fumbling around in a dark room until you find the light switch. At which point you can see the doorway leading to the next dark room.
That is how science seems to work as a whole. What worries me is that the market views the emergence of additional productive paradigm shifts in AI as only a matter of money. A normal scientific advancement plateau for another five years in AI would be a short-term disaster for the stock market and economy.
This is actually a lazy approach as you describe it. Instead, what is needed is an elegant and simple approach that is 99% of the way there out of the gate. Soon as you start doing statistical tweaking and overfitting models, you are not approaching a solution.
In a way yes. For models in physics that should make you suspicious, since most of our famous and useful models found are simple and accurate. However, in general intelligence or even multimodal pattern matching there’s no guarantee there’s an elegant architecture at the core. Elegant models in social sciences like economics, sociology and even fields like biology tend to be hilariously off.
I've singled out The Economist because I used to read them until not that long ago so that I can confirm first-hand that they also use that rhetoric (but can't be bothered to look for an online source right now).
Later edit: A X [2] post pointing to an Economist article [3] that does just that, but, as I said, the examples are too numerous, just purchase a Economist issue and go through their China section, you'll see it right there
