T
Towards AI
pub.towardsai.net · Group blog · on HN since 2021
Towards AI is a Medium publication where practitioners and writers share articles on machine learning, LLMs, and AI applications.
What HN saysHN links it for AI topics like Stable Diffusion image compression, Alpaca, and a piece on AI use causing trouble in the legal and medical fields.
83HN people shared it
18linked it in comments
83stories submitted
961story points
3 months agolast shared
2021people per year2026
Most-shared posts
- 1Stable Diffusion Based Image Compression8 people · 5 points · first shared 2022 · discussion on HN
- 2COSP and USP: New methods to advance reasoning in LLMs2 people · 28 points · first shared 2024 · 4 comments on HN
- 3Advanced RAG Techniques: An Illustrated Overview2 people · 1 points · first shared 2024 · discussion on HN
- 4Court Asked for the LLM's Reasoning. The Company Had Nothing. $10M2 people · 1 points · first shared 2026 · discussion on HN
- 5Spent $15 in DALL·E 2 credits creating this AI image1 people · 460 points · first shared 2022 · 149 comments on HN
- 673% of AI startups are just prompt engineering1 people · 246 points · first shared 2025 · 205 comments on HN
- 7The difference between a=a+b and a+=b in Python1 people · 58 points · first shared 2021 · 58 comments on HN
- 8The Rise of Software 2.01 people · 25 points · first shared 2021 · discussion on HN
- 9We spent 47k running AI agents in production1 people · 9 points · first shared 2025 · 7 comments on HN
- 10Comparing FastAPI and BentoML1 people · 9 points · first shared 2023 · 1 comments on HN
- 11AI Failed in the Face of Covid1 people · 8 points · first shared 2021 · discussion on HN
- 12KNN (K-Nearest Neighbors) Is Dead. Long Live ANN1 people · 5 points · first shared 2021 · discussion on HN
- 13Python lists Vs Tuple – In depth comparison1 people · 4 points · first shared 2021 · discussion on HN
- 14France's New AI Champion Scares Silicon Valley1 people · 3 points · first shared 2023 · 2 comments on HN
- 15Show HN: iPhone Orientation from Images and Sensors1 people · 3 points · first shared 2022 · 2 comments on HN
What HN people wrote when they linked it
anovv · 2024-04-08
Volga is an open-source feature calculation engine for real-time ML systems aimed to remove dependency on complex Spark/Flink setups and managed feature platforms like Tecton.ai, Fennel.ai, FeatureForm, Chalk.ai
Github - https://github.com/anovv/volga Medium - https://pub.towardsai.net/volga-open-source-data-engine-for-real-time-…
Would love to hear feedback, especially if you have experience building real-time feature calculation pipelines.
mgfist · 2023-03-23
> something comes out that kills GPT.
Can't even assume the LLM providers will be the big winners. LLMs can be replicated: https://pub.towardsai.net/meet-alpaca-stanford-universitys-instruction…
for much cheaper than they can be built.
isaacfrond · 2022-09-29
I'd be interested to see lossy text compression. That would also give much more scope for AI methods. It turned out that recently Stable Diffusion gives excellent compression
https://pub.towardsai.net/stable-diffusion-based-image-compresssion-6f…
Funny thing is, on a per pixel basis, Stable Diffusion and jpeg are just as good. Stable diffusion looks much better, though. The reason is that when Stable diffusion lacks the information it just invents what it expects to see there. If you zoom in on the background entire appartement buildings are invented, moved or disappeared.
So, one wonders what GPT-3 could do for text compression. On a per-character basis, I would not expect any miracles. On a generally, the same, kind-of, basis, I'd would expect something special.
jqpabc123 · 2026-04-22
As lawyers get burned by AI, they'll start to recognize the tremendous business opportunity it provides to burn others who unwisely employ it for any sort of "serious" work.
Here is a recent example from the medical industry:
https://pub.towardsai.net/the-air-gapped-chronicles-the-court-asked-fo…
Bottom line: Deferring to the "AI blackbox" (AI did it) won't fly in court and is a liability issue waiting to happen.
Dwedit · 2024-03-18
There was an earlier article (Sep 20, 2022) about using the Stable Diffusion VAE to perform image compression. Uses the VAE to change from pixel space to latent space, dithers the latent space down to 256 colors, then when it's time to decompress it, it de-noises that.
https://pub.towardsai.net/stable-diffusion-based-image-compresssion-6f…
HN discussion: https://news.ycombinator.com/item?id=32907494
ivan_ilin · 2024-02-08
My recent medium post on advanced RAG went kinda viral, but I'd like to see if there are some new techniques or setups to check out. Link to my medium post https://pub.towardsai.net/advanced-rag-techniques-an-illustrated-overv…
supriyo-biswas · 2024-02-05
In another adjacent thread, people are talking about the implications of a neural network conforming to the training data with some error margin with regards to copyright.
Many textbooks on information theory already call out the content-addressable nature of such networks[1], and they’re even used in applications like compression due to this purpose[2][3], and therefore it’s no surprise that the NYT prompting OpenAI models with a few paragraphs of their articles reproduced them nearly verbatim.
[1] https://www.inference.org.uk/itprnn/book.pdf
[2] https://bellard.org/nncp/
[3] https://pub.towardsai.net/stable-diffusion-based-image-compresssion-6f…
sgt101 · 2024-01-25
Google can be your friend brother (still).
https://pub.towardsai.net/inside-cosp-and-usp-google-research-new-meth…
zimpenfish · 2023-12-28
> I've never seen an excel sheet do anything like backpropagation.
Not strictly in the sense you mentioned (assuming that you mean "by themselves") but people may find [1] and [2] interesting.
[1] https://pub.towardsai.net/building-a-neural-network-with-backpropagati…
[2] https://towardsdatascience.com/demystifying-feed-forward-and-back-prop…
naasking · 2023-11-16
I understand the nuance, I think you're just wrong. LLMs augmented with memory are Turing complete [1], so they do process information. This distinction you're trying to draw is a mirage.
[1] https://pub.towardsai.net/llms-and-memory-is-definitely-all-you-need-g…
PaulHoule · 2023-05-09
I like the idea. Note that LLMs have some skill at decoding sequential dense vectors in the human brain
https://pub.towardsai.net/ais-mind-reading-revolution-how-gpt-llms-dec…
so why not have them decode sequential dense vectors of their own activations?
As for the majority scoring poorly, they suggest that most neurons won't have clear activation semantics so that is intrinsic to the task and you'd have to move to "decoding the semantics of neurons that fire as a group"
grumbel · 2023-04-29
It doesn't need to be lossless to be useful. A lossy LLM can just throw all the redundant stuff away and focus on the actual information in the text, instead of the text word by word. It wouldn't need to keep dozens of languages around for if it could translate on the fly. And it wouldn't need to keep a full text when it can just have the raw facts and generate the text on the fly.
Not sure what the state of the art is with LLM text compression, but with StableDiffusion compression ratios of up to 155:1 have been achieved[1] compared to 10:1 for JPEG.
[1] https://pub.towardsai.net/stable-diffusion-based-image-compresssion-6f…
timr · 2023-03-28
Yours is also just an opinion. But having a baseline understanding of what came before is important to having an informed opinion. To that end: GPT is not the first tech to combine "NLP" into a single model. For example, from 2019 [1].
This stuff is evolutionary.
[1] https://pub.towardsai.net/unified-language-model-pre-training-for-natu…
theptip · 2023-02-15
Look at OpenAssistant (https://github.com/LAION-AI/Open-Assistant); they are trying to give a chatbot the ability to trigger actions in other systems. I fleshed out a scenario in more detail here: https://news.ycombinator.com/item?id=34808674.
But in brief, the short-term evolution of LLMs is going to involve something like letting it `eval()` some code to take an action as part of a response to a prompt.
A recent paper, Toolformer: https://pub.towardsai.net/exploring-toolformer-meta-ai-new-transformer… which is training on a small set of hand-chosen tools, rather than `eval(<arbitrary code>)`, but hopefully it's clear that it's a very small step from the former to the latter.
systoll · 2022-12-21
There’s been some investigation in that direction, Eg: https://pub.towardsai.net/stable-diffusion-based-image-compresssion-6f…
It seems… dangerous to use this for lossy compression. JPEG artefacts look like blocks and blurriness. AI artefacts can be totally plausible changes to whatever is depicted.
But… it wouldn’t be the first time that has happened https://www.theregister.com/2013/08/06/xerox_copier_flaw_means_dodgy_n…
sharemywin · 2022-12-12
here was an article on using latent embeddings for compression. might be useful.
https://pub.towardsai.net/stable-diffusion-based-image-compresssion-6f…
SuperCuber · 2022-09-29