panarky

43 ms·

661.
by panarky 3y ago
The article says they changed auction rules that had the ultimate effect of increasing ad prices for some advertisers, at certain times, of between 5% and 10%. Even if that's true, it's still an auction, with real competition and
662.
by panarky 3y ago
"Bard conversation" can be different from "private information in Drive and Gmail".
663.
by panarky 3y ago
Agreed, many issues with Bard, though I've had a better experience working with the underlying PaLM 2 directly. Even given all of its faults, have you found a better LLM than Bard that can incorporate current events? Because for my pur
664.
by panarky 3y ago
>>>>>> Why does the comment field exist if not to write a comment? Maybe the design of the site should reflect the policy, instead of ambiguity that results in polluting the entire first page with meta-commentary about H
665.
by panarky 3y ago
What if it immediately followed a streak of 36 days above 110F?
666.
by panarky 3y ago
How would you re-write the statute so that it would apply to this?
667.
by panarky 3y ago
According to the literal text of the statute quoted there, the definition of "machinegun" hinges on a "single function of the trigger". If the shoestring achieves rapid fire by causing multiple functions of the trigger,
668.
by panarky 3y ago
> I've been testing every new LLM with that question We should pay more attention to data contamination when using popular prompts for testing.
669.
by panarky 3y ago
The first unstated assumption is that similar vectors are relevant documents, and for many use cases that's just not true. Cosine similarity != relevance. So if your pipeline pulls 2 or 4 or 12 document chunks into the LLM's conte
670.
by panarky 3y ago
I've tested both π and g, and while they both work well, g results in far fewer disk full errors. I've heard c works even better, though I haven't tried it yet.
671.
by panarky 3y ago
Very curious to follow your journey through embedding search. If I want 100 close matches that match a filter, is it better to filter first then find vector similarity within that, or find 1000 similar vectors and then filter that subset?
672.
by panarky 3y ago
> refuge for Christian sects ... Something like 25% of European settlers were fleeing religious persecution. The vast majority were economic migrants. Religious migrants: 20,000 (Massachusetts Bay) + 3,000 (Plymouth) + 10,000 (half of
673.
by panarky 3y ago
Tesla claims they've solved both the intermittancy of renewable energy and bursty demand patterns with grid-scale solutions like Megapack. There's no excuse for running diesel generators.
674.
by panarky 3y ago
I'll take that bet. I'm guessing something electricity-related during a flood, or CO poisoning from a grill or heater inside a tent or trailer. Also, one death per year is typical. -------- 2010: A man died in his tent from carbon
675.
by panarky 3y ago
Most years see one death on the playa. 2010: A man died in his tent from carbon monoxide poisoning due to a propane heater. 2011: A woman died from injuries sustained in a tent fire. 2012: A man was found dead in his tent; the cause was not
676.
by panarky 3y ago
I'm experiencing a bizarre sense of unreality reading comments on a story about Amazon killing EC2-Classic, where the first line is "Retiring services isn’t something we do", while having just read yet another email from Amaz
677.
Polars vs. Pandas: What’s the Difference?(blog.jetbrains.com)
4 points by panarky 3y ago|0 comments
678.
by panarky 3y ago
Those percentage increases seem to be lower than the rate of inflation at that time, which would mean the real value of imports and exports was flat or lower.
679.
by panarky 3y ago
PaLM 2 is the underlying model for more than two dozen Google products, such as MedPaLM 2. Each product fine-tunes their own weights on top of the underlying model, but it seems like that fine-tuning is more about style than substance. For
680.
Tune PaLM 2 with your own RLHF training data(github.com)
17 points by panarky 3y ago|4 comments
681.
by panarky 3y ago
What safeguards prevent the LLM from generating Cartesian products?
682.
by panarky 3y ago
"Imagine you’re a Google PR writer and you get an email at 5 PM from your boss asking for a blog post about Duet AI for release to the public at 8 AM tomorrow ..."
683.
Cloud TPU v5e and Nvidia H100 A3 GA(cloud.google.com)
16 points by panarky 3y ago|1 comments
684.
AI Anywhere with Google Distributed Cloud(cloud.google.com)
2 points by panarky 3y ago|0 comments
685.
Expanding Duet AI Across Google Cloud(cloud.google.com)
3 points by panarky 3y ago|0 comments
686.
by panarky 3y ago
Anthropic's Claude 2 32k context window for PaLM 2 Adapter tuning and RLHF for LLaMA 2
687.
New models, tooling for Vertex AI generative AI(cloud.google.com)
10 points by panarky 3y ago|5 comments
688.
by panarky 3y ago
It's so awesome to learn a new term like "macaronic language". https://en.wikipedia.org/wiki/Macaronic_language I finally have the right term to describe the warning signs from 1960's-era mainframes
689.
by panarky 3y ago
NEVER CREATE CARTESIAN PRODUCTS For example: SELECT clicks.session_id, clicks.page_url, sessions.user_id FROM clicks, sessions; (This context is guaranteed to generate lots of cartesian products.)
690.
by panarky 3y ago
If I'd tried to start my kids on how to install software on an operating system, and how to write bash and SQL, they would have lost interest instantaneously. I found the gateway drug was a robot that could do simple things, then modif

More ›