ttul

21 ms·

331.
by ttul 2y ago
See my comment above. This is a misunderstanding of how the executive branch works. Once Congress has enacted a statute and the _President_ has signed it into law, the executive branch MUST enforce it. An executive order cannot override or
332.
by ttul 2y ago
Here’s the law: https://www.congress.gov/bill/118th-congress/house-bill/7521... No, Trump can’t legally postpone or give reprieve to TikTok. The time has passed for that. Once Congress has enacted a statute a
333.
by ttul 2y ago
It wasn’t quite as gauche as I made it sound in my comment. The fishing invitation was extended to a few customers and was an official AWS sponsored event.
334.
by ttul 2y ago
I mean, OpenAI not only tested the model, they literally trained the model. Training a model involves developing evaluations for the model. It’s a gargantuan effort. I’m fairly certain that OpenAI is the authority on how to prompt o1.
335.
by ttul 2y ago
FWIW: OpenAI provides advice on how to prompt o1 ( https://platform.openai.com/docs/guides/reasoning/advice-on-... ). Their first bit of advice is to, “Keep prompts simple and direct: The models excel at unders
336.
by ttul 2y ago
My AWS account manager took me fishing. That’s what you get for a >$1M/yr spend. I don’t sense they would roll in mud with me, which is kind of incredible. I wonder how much you need to spend to get into mud rolling territory?
337.
by ttul 2y ago
This is big news.
338.
by ttul 2y ago
This paper was written by a very small team at Google. It strikes me as similar in that regard to the original transformers paper. If this technique scales well, Google is no doubt already exploiting it for their next generation models -- a
339.
by ttul 2y ago
No, TITANS introduces a new "learning memory" model that is actually trained at test-time based on the input to help remember the right bits from the input to improve how the separate sequence-to-sequence model generates tokens. T
340.
by ttul 2y ago
Metaculus - this is the median prediction.
341.
MiniMax-01: Scaling Foundation Models with Lightning Attention(arxiv.org)
3 points by ttul 2y ago|1 comments
342.
by ttul 2y ago
“ We introduce MiniMax-01 series, including MiniMax-Text-01 and MiniMax-VL-01, which are comparable to top-tier models while offering superior capabilities in processing longer contexts. The core lies in lightning attention and its efficien
343.
by ttul 2y ago
Amazon had an insane number of people working on just the alarms feature in Alexa when they interviewed me for a position years ago. They had entire teams devoted to the tiniest edge case within the realm of scheduling things with Alexa. Th
344.
by ttul 2y ago
Given that there may be a 25% chance that China invades Taiwan by 2030, having the ability to package SOTA chips in the US by 2027 seems "soon enough".
345.
by ttul 2y ago
I run a cybersecurity company and I’ve had drinks with Krebs at various events over the years. He’s the real deal, digging up dirt on the people who ruin everything for everyone and risking his life in the process for a minuscule payoff. I
346.
by ttul 2y ago
I bought a Lamba Labs workstation with a 4090 last year. I guess I’m buying one of these things now because the Lambda workstation just became a relic…
347.
by ttul 2y ago
Organizations should be spear phishing their employees using LLMs to test defenses and identify gaps in security awareness training and processes.
348.
by ttul 2y ago
Also: Odd Lots.
349.
by ttul 2y ago
Thanks for your extensive explanation!
350.
by ttul 2y ago
OpenAI’s credibility is central to its business: overstating capabilities risks public blowback, loss of trust, and regulatory scrutiny. As a result, it is unlikely that OpenAI would knowingly lie about its models. They have much stronger i
351.
by ttul 2y ago
No, this all happens inside the model. I suppose it’s possible that the hidden layers of one model could be sent to another model. But the second model would need to be trained to understand the meaning of the hidden layer’s outputs. You co
352.
by ttul 2y ago
Output tokens are still generated, otherwise the model wouldn’t know when to stop being in latent space mode. The <eot> token emerges as the top token at the output layer when it’s time to switch back.
353.
by ttul 2y ago
Sign up for MailChannels API and I’ll make it free for you.
354.
by ttul 2y ago
I’d love to tune in for updates, but the subscribe button says, “ Failed to send verification email.” This is so cool. Would love to follow along.
355.
by ttul 2y ago
TL;DR: Meta started with a pre-trained language model. They then fine-tuned it on step-by-step reasoning examples as you would do if you wanted your model to become particularly good at chain of thought reasoning. However, they also introdu
356.
by ttul 2y ago
It would indeed be foolish to pay an invoice to a company that has very publicly declared insolvency. Many if not most contracts I have ever signed contain a provision voiding the obligations of the parties if either party is insolvent.
357.
by ttul 2y ago
I’d like to clarify - and ought to have done above - that I’m not saying you’re lying in your Tweets. Just that the story of any business failure is often more nuanced than “investors are bad”. And that story may be revealed in court filing
358.
by ttul 2y ago
Someone should build an engineer-matchmaking service with a fresh revenue model that can challenge LinkedIn, whose platform is now so heavily optimized for profit-taking that it leaves both employer and job searcher unsatisfied. Using LLMs
359.
by ttul 2y ago
I advise caution in believing his statements to be true, particularly about Bench’s board and investors. The truth will live in their financial statements, which no doubt will be revealed in bankruptcy proceedings in the coming months.
360.
by ttul 2y ago
I’m sure all of this will be played out in spectacular fashion in the courts of British Columbia in the coming months. Directors in BC are liable for unpaid wages to employees. No doubt they pulled the rug just in time to ensure people got

More ›