fsndz

9 ms·

91.
by fsndz 2y ago
frankly ollama + Deepseek is all you need to win with open source AI. I will do some experiments today and add it to my initial blogpost. https://medium.com/thoughts-on-machine-learning/deepseek-is-...
92.
by fsndz 2y ago
now openai has no other choice than shipping a cheaper version of o1 and o3. The alternative is everyone using r1 (self hosted or via openrouter, nebius AI, together AI and co)
93.
by fsndz 2y ago
yes o3 is better, but I would argue it is not yet clear for which cases it is absolutely crucial to use o3 instead of o1.
94.
by fsndz 2y ago
indeed there is no moat. Open source will win !
95.
by fsndz 2y ago
this means we are going to get o3 level open source models in a few months. So exciting !
96.
by fsndz 2y ago
The ratio is probably 4/5 for YC Startups
97.
by fsndz 2y ago
In his latest blog post, titled “Reflections”, Sam Altman, the CEO of OpenAI, wrote that they are now confident about their ability to build AGI (Artificial General Intelligence). That’s a pretty serious statement to make, and I don’t think
98.
by fsndz 2y ago
this feels like you are talking about langchain haha: https://medium.com/@fsndzomga/jailbreaking-langchain-ad9415a...
99.
by fsndz 2y ago
2 years in we are still going and it is still becoming more and more useful. Keep in mind, things like multimodality are still in early stages so improvements are coming. It probably won't lead to AGI but still, these tools are doomed
100.
by fsndz 2y ago
I have been attempting to do this and failed. Good implementation, but still fails in a lot of cases like my implementation. For example fails for this: https://github.com/stanfordnlp/dspy
101.
by fsndz 2y ago
Interesting article. My main criticism is that, given ChatGPT is already used by hundreds of millions of people every day, it's difficult to argue that current AI is a dead end. It has its flaws, but it is already useful in human-in-th
102.
by fsndz 2y ago
yes indeed. But I see more and more people arguing against the very possibility of AGI. Some people say statistical models will always have a margin of error and as such will have some form of reliability issues: https://open.sub
103.
by fsndz 2y ago
That's exactly what happens when AI realism fades from the picture: inflated expectations followed by disappointments. We need more realistic visions, and we need them fast: https://open.substack.com/pub/transition
104.
by fsndz 2y ago
I get the excitement, but folks, this is a model that excels only in things like software engineering/math. They basically used reinforcement learning to train the model to better remember which pattern to use to solve specific problem
105.
by fsndz 2y ago
that's a false choice because the SaaS startup will often be willing to modify the product to improve it for customers
106.
Show HN: Let a Vision Model Roast You(qwenroast.vercel.app)
1 points by fsndz 2y ago|0 comments
107.
by fsndz 2y ago
I don't get the comparison. are you comparing a SaaS that is doing 70% of the job and a shitty streamlit POC that no more than 3 people can use in parallel ??
108.
by fsndz 2y ago
haha this is why I love HN
109.
by fsndz 2y ago
True and what I find funny is that the same companies are often willing to pay a lot of money to consultants that will deliver a shitty streamlit POC. Like they will literally pay consultants the equivalent of 4 years of new SaaS cost to ge
110.
by fsndz 2y ago
patching all the way down. this is no way to reach AGI. we need more AI realism, the hype is too much right now: https://open.substack.com/pub/transitions/p/why-ai-realism-m...
111.
by fsndz 2y ago
That's why AI realism is deeply needed now: https://transitions.substack.com/p/why-ai-realism-matters?r=...
112.
by fsndz 2y ago
In a recent article about the capabilities and implications of AI, particularly large language models (LLMs), Casey Newton framed the discussion around two extremes: those who claim AI is "fake and sucks" and those who insist AI i
113.
by fsndz 2y ago
I find it bizarre that people often criticize LLMs’ capabilities by stating they don’t exhibit human-level performance because they sometimes fail, when occasionally failing is, in fact, a quintessentially human trait. Moreover, how do we r
114.
by fsndz 2y ago
AI engineers just use "vibes" currently haha
115.
by fsndz 2y ago
I wonder if there will ever be another programming language on top of the most high level languages we currently have (like to python what python was for C)
116.
by fsndz 2y ago
Synthetic data generation is one of the lesser-discussed use cases of LLMs, yet it holds significant potential. I’ve created a rudimentary interface to test this concept using open-source LLMs available on Nebius AI Studio.
117.
by fsndz 2y ago
so your solution is a niche language nobody uses ?
118.
by fsndz 2y ago
same is true for AI music, AI art, AI articles to some extent. most communication is about human connection. you remove it and the communication becomes worthless. AI is great for communication that has less personal involvement like drafti
119.
by fsndz 2y ago
I feel like HumanLayer is a great idea, but decision fatigue and bystander effects could pose challenges. If people are overloaded with approvals or don't feel ownership over what they're verifying, the quality of oversight might
120.
by fsndz 2y ago
I stopped at: "causal sequence of “thoughts” "

More ›