jsemrau

7 ms·

61.
by jsemrau 1y ago
>Maybe hedge funds or some sort of financial markets? I'd think that HFT is already mature and doesn't really benefit from this type of model.
62.
by jsemrau 1y ago
Yes, that's commonly referred to as the Exploration-Exploitation Dilemma. Should the agent go deep or wide? https://en.wikipedia.org/wiki/Exploration%E2%80%93exploitati...
63.
by jsemrau 1y ago
If it were only tool use, then it would be the same as a lambda function.
64.
by jsemrau 1y ago
Agents are more than that. Agents, besides tool use, also have memory, can plan work towards a goal, and can, through an iterative process (Reflect - Act), validate if they are on the right track.
65.
by jsemrau 1y ago
You can use a lower-end GPU (like the RTX 3060), which also uses less energy. But you are right, you won't be encountering model API costs when running it locally.
66.
by jsemrau 1y ago
It's not unprecedented, but I question if it is the right move while the industry is experiencing unprecedented growth. "The Global Data Center Chip Market size is expected to be worth around USD 57.9 Billion by 2033, from USD 14.
67.
by jsemrau 1y ago
Once you are working with local LLMs you quickly run into CUDA Out of Memory error. Managing input context input sizes in prompts is really critical. Also keeps cost down.
68.
by jsemrau 1y ago
Cognition also works with Goldman Sachs now.
69.
by jsemrau 1y ago
Try it out? https://deepmind.google/models/gemini-diffusion/
70.
by jsemrau 1y ago
Tesla was/is? Luminar's biggest customer [1]. So they used it for something, likely ground-truth validation [2] [1] https://www.reuters.com/technology/luminar-says-tesla-is-big... [2] https://www.t
71.
by jsemrau 1y ago
I still hold the belief that the removal of Radar and LIDAR in 2021 was only detrimental to FSD's performance.
72.
by jsemrau 1y ago
"The standard for "agents" is that tools run in sequence" I don't think that this correct. Agents benefit is that they can use tools on the fly. Ideally the right tool at the right time. I.e., Which number is bigger
73.
by jsemrau 1y ago
(1) Running multiple agents is expensive, decreasing RoI. My DeepSearch agent for stocks uses 6 agents, and each query costs about 2 USD. (2) Multi-agent orchestration is difficult to control. (3) The more capable the model, the lower the n
74.
by jsemrau 1y ago
"your efforts will lead to a discrimination of 20 years old females" I'd think that this is an extremely far-fetched example that fails at basic logic. Just because a very specific scenario will be flagged does not mean that
75.
by jsemrau 1y ago
>The goal is to avoid penalizing people for their skin color [...] That's not correct. The goal is to identify and flag fraud cases. If one group has a higher likelihood to perform that, then this will show up in the data. The sol
76.
by jsemrau 1y ago
You can't take that boat across that ocean.
77.
by jsemrau 1y ago
Most agent frameworks implement a ReAct (Reflect->Act) pattern in the form of a loop that "reasons" until a "final_answer" has been found. Most framworks also implement some form of session cache (scratchpad), tool u
78.
by jsemrau 1y ago
"But here's the kicker"
79.
by jsemrau 1y ago
Up until a few weeks ago, my LinkedIn seemed to become better because of AI, but now it seems everything is lazy AI slop. We meatbags are great pattern recognizers. Here is a list of my current triggers: "The twist?", "Then
80.
by jsemrau 1y ago
Is that really a product? I think it should be solved through workflow and policies rather than providing this to a 3rd party provider. But I might be wrong. [1] https://jdsemrau.substack.com/p/memory-and-context
81.
by jsemrau 1y ago
I had good results with Nemo + xTTS_v2 https://docs.nvidia.com/nemo-framework/user-guide/latest/nem... https://huggingface.co/coqui/XTTS-v2
82.
by jsemrau 1y ago
I have a signal tracer that evaluates unusual trading volumes. Given those signals, my local agent receives news items through API to make an assessment what happens. This helps me tremendously. If I would do this through a remote app, I&#x
83.
by jsemrau 1y ago
The tighter the scope/validation loop, the closer the "agent" gets to a more narrow business case of machine learning. These more traditional cases are in comparison significantly cheaper to implement and maintain. If you con
84.
by jsemrau 1y ago
I still haven't found what an actual working product in AI Agents could be and write about my journey into capabilities, frameworks, and restrictions here: https://jdsemrau.substack.com/ Initially I thought there is a
85.
by jsemrau 1y ago
>Harvard and Yale didn't hire the right lobbyists. I don't think it's as simple as this. To my knowledge, Dr. Sian Leah Beilock handled the protests of the past 2 years much better than their counterparts.
86.
by jsemrau 1y ago
"They do not have a plan" Not necessarily correct if you consider agent architectures where one LLM would come up with a plan and another LLM executes the provided plan. This is already existing.
87.
Jungle Ventures-, QIA-backed unicorn Builder.ai enters insolvency proceedings(dealstreetasia.com)
1 points by jsemrau 1y ago|1 comments
88.
by jsemrau 1y ago
https://www.dealstreetasia.com/stories/builder-ai-collapses-...
89.
by jsemrau 1y ago
When I did my research on reasoning strategic games in 4x4 tic tac toe boards, LLMs with given nominal parameters and low temperature would still show variance in their assessment of the situation.
90.
by jsemrau 1y ago
If the question is to understand the default training/bias then this approach does make sense, though. For most people LLMs are black box models and this is one way to understand their bias. That said, I'd argue that most LLMs are

More ›