andrewdb

7 ms·

1.
by andrewdb 7mo ago
Why do we call them GPUs these days? Most GPUs, sitting in racks in datacenters, aren't "processing graphics" anyhow.
2.
by andrewdb 7mo ago
Sure, there is a problem with slop AI PRs _now_ . That will not remain true for infinity. What happens when the AI PRs aren't slop?
3.
by andrewdb 7mo ago
We are getting to a point where AI will be able to construct sound arguments in prose. They will make logical sense. Dismissing them only because of their origin is fallacious thinking. Conclusion: Dismissing arguments solely because they a
4.
by andrewdb 7mo ago
If the PR had been proposed by a human, but it was 100% identical to the output generated by the bot, would it have been accepted?
5.
by andrewdb 7mo ago
So when will multiple Waymo cars communicate input data to one another to avoid the blind spots? This would give the ability to see things other cars cannot see as well.
6.
The death of software development(mike.tech)
33 points by andrewdb 8mo ago|25 comments
7.
by andrewdb 2y ago
One way to slightly mitigate the difficulties of nuance in language when translating to formal arguments is to attwmpt to always steelman the argument. Afford it all the guarded language and nuance you can, and then formalize in premises an
8.
by andrewdb 2y ago
A prompt that I like to use for this: --- Intake the following block of text and then formulate it as a steelmanned deductive argument. Use the format of premises and conclusion. After the argument, list possible fallacies in the argument.
9.
by andrewdb 2y ago
Not just chip exports. It also limits model weights. From the article: In addition to the semiconductor controls, the new rules also limit the export of closed AI model weights, which are the numerical parameters that software uses to proce
10.
by andrewdb 2y ago
In large part, yes
11.
by andrewdb 2y ago
Steel-manned Deductive Argument Premise 1: AI (specifically, statistical modeling based on hidden layer neural networks) has been increasingly integrated into various technological products and services. Premise 2: Major companies like Micr
12.
by andrewdb 2y ago
Those are good suggestions. I will use some of them! It is also interesting to go back and forth with the model, asking it to mitigate fallacies listed, and then re-check for fallacies, then mitigate again, etc, etc. I have found that a wor
13.
by andrewdb 2y ago
I have found the below to be a good starting point for formulating text into classical formulated arguments. Intake the following block of text and then formulate it as a steelmanned deductive argument. Use the format of premises and conclu
14.
by andrewdb 2y ago
If the nutritional value of a plant's output is governed by its genetics, then this should be a solvable problem. Using the boosted method, breed a high nutritional value plant with a high output plant.
15.
by andrewdb 3y ago
If only this Dockerfile were real. It would greatly help app developers and publishers: FROM apple/mac-os-slim:latest ... Please, Apple, please let your developers use more virtualization or containerization.
16.
by andrewdb 3y ago
An over-simplified approach that would be interesting to explore: 1. Given a collection of logical arguments, figure out how to assign a hash to each argument. 2. Build a Merkle tree representing the argument's hashes, nesting accordin
17.
by andrewdb 3y ago
Steel-manned deductive argument Premise 1: The consumption of energy and natural resources required for AI infrastructure, particularly if it continues to grow, is massive and potentially unsustainable. Premise 2: The economic investment of
18.
by andrewdb 3y ago
Possible fallacies in the post's argument Slippery Slope: The argument may overstate the direct line from investment in AI to catastrophic outcomes like global financial depression or war over resources. Appeal to Fear: Highlighting ex
19.
by andrewdb 3y ago
Grats on the launch! Would be very interested to see this working with GCP Cloud Run.
20.
by andrewdb 3y ago
A good test of LLMs: > Give me a list of 5 words having 5 syllables each Mistral 8x7B and GPT 4 both choke on this frequently.
21.
by andrewdb 3y ago
"I don't play craps. I'm a dice engineer!" - prompt engineers, satirically
22.
by andrewdb 4y ago
yes.
23.
by andrewdb 4y ago
I assume you're talking about mDAU. This is what Twitter says, essentially: "We used internal data to produce a judgement-driven estimate that something is X, but we could be wrong, and it could be higher than X". How is some
24.
by andrewdb 4y ago
Not only is the trial schedule not changed, but Chancellor McCormick also made it clear she had little patience for expansive discovery for the whistleblower points. She is only allowing targeted, incremental discovery for these new additio
25.
by andrewdb 4y ago
Musk has the burden of proof to show fraud or MAE (material adverse effect). For fraud, he needs to prove scienter as well as reliance. High bar, and undermined by his own tweets and podcast appearances. For MAE, it requires Musk to show Tw
26.
by andrewdb 4y ago
Premise 1: Showing ads to bots is generally not good for business. Premise 2: Showing ads to humans is generally good for business. Premise 3: A cohort can be created to hold humans (called mDAU). Premise 4: Only show ads to the mDAU cohort
27.
by andrewdb 4y ago
Twitter's method for counting bots, according to Parag: 1. Sample random X accounts from known mDAU accounts 2. Use human analysis to manually review sampled accounts, using public and private data and metadata 3. Bucketize account acc
28.
by andrewdb 5y ago
Relevant: Brandolini's Law "The amount of energy needed to refute misinformation is an order of magnitude larger than is needed to produce it." It can take ~15 minutes to create a Twitter post containing misinformation, but a
29.
by andrewdb 6y ago
You can set minimum instances now ( well - currently Preview). https://cloud.google.com/run/docs/configuring/min-instances
30.
by andrewdb 6y ago
Re: worker processes Have a look at GCP EventArc. It's a new way to trigger cloud run instances. You could create an EventArc trigger that wakes up a worker image, or some such. Basically,EventArc enables serverless event driven design

More ›