Concentration of power in AI is a risk, not a solution
The answer to concentration of power in AI is not openness at all costs, but a serious research commons with the resources to compete.
Andy Konwinski
Notes on AI, RL, systems, etc.
Latest posts
K Prize round one closed; what’s next
K Prize round one submissions are closed
The first Laude Salon: on researchers shaping AI’s impact
A conversation about how researchers can shape the AI agenda
Why I built the Konwinski Prize
Why I launched the K Prize, aka, why want to give $1M to the first open source AI that gets 90% on contamination-free SWE-bench.
Laude Ventures
A few years ago, I co-founded a small fund called CSGV (CS Grad Ventures) that took money from — and invested it in — Computer Science PhDs and Professors. Our vision was to democratize Venture Capital back to the researchers themselves. To succeed we would need to find the next Databricks.
Introducing Headless Terminal
Introducing Headless Terminal (ht) - making terminals easy for LLMs to use.
LLMs are like toddlers
I think my 3yr old and Llama 3 were separated at birth. The way training my daughter involves (verbally) collecting a dataset of thousands of examples. The way she sometimes needs only one example, sometimes dozens. The way she hallucinates (e.g., while learning to count, she skipped “13” for months). However, thanks to my biological imperitive as a dad, I find it deeply satisfying to spend…
AI benchmarks should be like unit tests
Today’s go-to LLM benchmarks incentivize model makers to improve marginal performance on tasks that AIs (i.e., a model or model+memory+tools+etc.) have already mastered at superhuman-levels like general Q&A, fact recall, summarization, simple few-step reasoning, etc. In contrast, my favorite benchmarks (like SWE-bench) are more like unit tests in that they test a skill which AIs still haven’t…
Focus on the failures
When playing LLMs, it is tempting to only share and discuss the prompts that cause the model to return something interesting that demonstrates the models reasoning capabilities or the knowledge it has memorized.
List of Autonomous LLM Agent Projects
Last Updated: April 19, 2023