Sign in

Nino

@ninohardt.bsky.social
66 followers 333 following 68 posts

Marketing scientist, market researcher, choice modeler. web: ninohardt.com

PostsRepliesMedia
Nino @ninohardt.bsky.social · 24/12/2025
www.sca.isr.umich.edu hmm
sca.isr.umich.edu
Surveys of Consumers
000
Nino @ninohardt.bsky.social · 17/11/2025
Hypothesis: AI enables more rational decision making by consumers. Evidence, exhibit 1:
000
Nino @ninohardt.bsky.social · 13/11/2025
I guess it's either "latest" or "legacy"
000
Nino @ninohardt.bsky.social · 21/05/2025
Waitlist. How will it compare to chat.inceptionlabs.ai?
chat.inceptionlabs.ai
Mercury Coder
Mercury Coder
000
Nino @ninohardt.bsky.social · 02/04/2025
Parallels to reward hacking issue?
010
Nino @ninohardt.bsky.social · 19/03/2025
Based on the image, it works
010
Nino @ninohardt.bsky.social · 19/03/2025
@philschmid.bsky.social saw you tweet (_philschmid/status/1902403286902178000) saying "Understands the text and visuals like Charts in the PDF", but I have doubts.
000
Nino @ninohardt.bsky.social · 19/03/2025
PDF support in Gemini is great, but reading numbers off a line chart isn't quite there yet.
220
Nino @ninohardt.bsky.social · 09/03/2025
@rao2z.bsky.social introduced the term "prompt augmentation". The holy grail would be a skolem function that finds the best prompt augmentation for an optimal output, whatever it may look like (it might not even make sense to humans).
010
Nino @ninohardt.bsky.social · 06/03/2025
Turns out it isn't quite so straightforward: arxiv.org/abs/2502.21098
arxiv.org
Re-evaluating Theory of Mind evaluation in large language models
The question of whether large language models (LLMs) possess Theory of Mind (ToM) -- often defined as the ability to reason about others' mental states -- has sparked significant scientific and public...
000
Nino @ninohardt.bsky.social · 05/03/2025
Lot's of talk about AGI, but no computer scientist involved. www.nytimes.com/2025/03/04/o...
nytimes.com
Opinion | The Government Knows A.G.I. Is Coming
The Biden administration’s A.I. adviser Ben Buchanan discusses how the U.S. government is preparing for artificial general intelligence — and all the challenges that remain.
010
Nino @ninohardt.bsky.social · 04/03/2025
Say what you want about OpenAI, but its UX doesn't frustrate me the same way as Anthropic Claude.
010
Nino @ninohardt.bsky.social · 02/03/2025
It's just my first impression. I asked it to act as a psychologist conducting an In-Depth-Interview (IDI) for a consumer good. It stayed in-character (great), but would get stuck in rabbit holes. Latency and EQ are great though.
100
Nino @ninohardt.bsky.social · 02/03/2025
It's sometimes trying a little too hard to be human, and I wonder how it'd do on instruction following.
100
Nino @ninohardt.bsky.social · 27/02/2025
What's the unit on the axis? # vibes? vibes/token?
000
Nino @ninohardt.bsky.social · 27/02/2025
For certain writing tasks, it is much better. Counting by the token doesn't do justice to the net tokens that actually end up in your final deliverable.
010
Nino @ninohardt.bsky.social · 27/02/2025
Which is better? 4.5? Seems more elegant.
000
Nino @ninohardt.bsky.social · 20/02/2025
So if unemployment↑ and inflation↑ what will Fed do?
000
Nino @ninohardt.bsky.social · 11/02/2025
Training for a new domain, with verifiable answers / solutions
000
Nino @ninohardt.bsky.social · 04/02/2025
@weightsbiases.bsky.social about to rename?
000
Nino @ninohardt.bsky.social · 31/01/2025
MarkText can help with cleaning up markdown, it's a good editor. But I just use Cursor, even on .md files in an Obsidian vault.
011
Nino @ninohardt.bsky.social · 31/01/2025
Notion? It's plain markdown, you can link between the markdown notes, and you can share.
110
Nino @ninohardt.bsky.social · 24/01/2025
Lex Fridman-esque
000
Nino @ninohardt.bsky.social · 23/01/2025
Citations on the Anthropic API is really nice. Maybe a gamechanger for real business applications. We'll see. Need to test it more. www.anthropic.com/news/introdu...
anthropic.com
Introducing Citations on the Anthropic API
Today, we're launching Citations, a new API feature that lets Claude ground its answers in source documents.
000
Nino @ninohardt.bsky.social · 23/01/2025
In consumer good context, 10.1287/mksc.2022.1371 show that generational share gap is mostly due to experience, and less due to intrinsic differences.
100
Nino @ninohardt.bsky.social · 23/01/2025
Yes, like that. Maybe you could categorize the controls.
000
Nino @ninohardt.bsky.social · 23/01/2025
With enough controls, can you make the cohort effect vanish?
100
Nino @ninohardt.bsky.social · 22/01/2025
I understand the sentiment, but can we just all stop anthropomorphizing AI? It's not helpful.
110
Nino @ninohardt.bsky.social · 22/01/2025
"applying theory of mind" sounds odd indeed. Just what do cognitive scientists make of that example? And what are the latest ToM & LLM papers worth reading?
100
Nino @ninohardt.bsky.social · 20/01/2025
"summary" is idiosyncratic yet people pretend it's not
010
Nino @ninohardt.bsky.social · 20/01/2025
In written, it works as expected.
000
Nino @ninohardt.bsky.social · 20/01/2025
Language detection in OpenAI AVM. Why does it recognize my Japanese but then proceeds to speak Spanish?
100
Nino @ninohardt.bsky.social · 19/01/2025
Google's AI Studio let's me paste long textual contexts and then engage in a live discussion. OpenAI's AVM doesn't let me do that.
000
Nino @ninohardt.bsky.social · 07/01/2025
NVIDIA reported INT8 for Jetson Orin, now FP4 for Digits. Who are they kidding?
110
Nino @ninohardt.bsky.social · 07/01/2025
FP4 ....
110
Nino @ninohardt.bsky.social · 30/12/2024
"good prompts for your data" is the new "let's train on the test set"
100
Nino @ninohardt.bsky.social · 29/12/2024
"democratization" ... with a censored model. Oh, the irony.
110
Nino @ninohardt.bsky.social · 26/12/2024
Interesting: when you ask for illegal drug recipes, deepseek behaves more like other models, it just writes out that it won't help. It's a different mechanism.
010
Nino @ninohardt.bsky.social · 26/12/2024
On a number of predictably "sensitive" topics, the model first generates and then abruptly deletes already generated text to say "beyond scope" or something.
110
Nino @ninohardt.bsky.social · 23/12/2024
What is "most"? >50%? And what is the list of "tasks"?
000
Nino @ninohardt.bsky.social · 20/12/2024
quoting from that page "Passing ARC-AGI does not equate to achieving AGI, and, as a matter of fact, I don't think o3 is AGI yet. o3 still fails on some very easy tasks, indicating fundamental differences with human intelligence."
000
Nino @ninohardt.bsky.social · 20/12/2024
Instead of just quoting "breakthrough", maybe also quote "Passing ARC-AGI does not equate to achieving AGI, and, as a matter of fact, I don't think o3 is AGI yet."
000
Nino @ninohardt.bsky.social · 19/12/2024
ChatGPT "working with" is nice, but feels incomplete. Unless it actually adds/changes itself, it doesn't add much value. And while AVM is nice, it failed to directly show me the content to add. Feels like a lot of work ahead for the team.
000
Nino @ninohardt.bsky.social · 18/12/2024
That was their intention, I guess. Well, it's still a useful device for edge applications, but I dislike that nvidia sometimes omits those kind of details ...
000
Nino @ninohardt.bsky.social · 18/12/2024
*INT8
110
Nino @ninohardt.bsky.social · 17/12/2024
veo veo ... que ves
000
Nino @ninohardt.bsky.social · 15/12/2024
Asked ChatGPT AVM video to help solve rubrics cube, and ... Doesn't help much. On that note, while it's easy to share prompts, sharing video/voice interactions isn't convenient.
000
Nino @ninohardt.bsky.social · 14/12/2024
The number of humans on the planet is at an ATH. They'll come up with some actually new things. It may take a little longer than the average time between new models on HF, but still doesn't mean we are at "peak" data.
000
Nino @ninohardt.bsky.social · 14/12/2024
Isn't that kinda insulting to anyone writing a book (or a paper, or ...) today? The data they generate doesn't count?
100
Nino @ninohardt.bsky.social · 13/12/2024
Got Video on AVM and Gemini-2.0-flash Live, and both are nice. Not sure which one is better yet.
000