Working on merging symbolic and subsymbolic AI.
Germany
Joined October 2010
- Tweets1.6K
- Following1.6K
- Followers403
- Likes1.8K
Kurt Uwe Stoll retweeted
The elegance of ML is the elegance of biology, not the elegance of math or physics.
Simple gradient descent creates mind-boggling structure and behavior, just as evolution creates the awe inspiring complexity of nature.
Kurt Uwe Stoll retweeted
🏡 I asked GPT4 to imagine if nomadlist.com would start coworking spaces + nomad bases around the world and how it'd look:
1) A renovated villa in Portugal transformed into a coworking space, overlooking the Atlantic Ocean.
2) A coworking space in Bali that blends traditional Balinese architecture with contemporary design, surrounded by rice fields.
3) An urban high-rise coworking space in Bangkok with a rooftop area.
4) An illustration showcasing a sustainable coworking space with solar panels, rainwater harvesting systems, and a community area surrounded by greenery.
Kurt Uwe Stoll retweeted
AgentTuning: Enabling Generalized Agent Abilities for LLMs
Presents AgentLM-70B, which performs comparably to GPT-3.5-turbo on unseen agent tasks, demonstrating generalized agent capabilities
arxiv.org/abs/2310.12823
Kurt Uwe Stoll retweeted
I burned in🔥2000$ in finetuning so you don't have to.
I fine-tuned models with @OpenAI and @anyscalecompute API endpoints with 50million tokens. Here are the results I wish I knew before getting into finetuning.
If you just want a quick snapshot, look at the figure. A longer explanation follows, explaining my findings.
I am not an expert and not deep into theory of AI models. I just want to get the BEST model performance at the CHEAPEST possible price for my USE-CASE. And quickly deploy that to prod.
I picked one specific simple USE-CASE. Summarizing text in a very specific tone, voice and a very specific structure.
Trained both models with close to 50M tokens (~37M words). In short,
- Anyscale costs 40X cheaper to finetune.
- Anyscale costs 56x cheaper to finetune.
Comparing the outputs, I get on par performance from llama-13b-fine-tuned as gpt-3.5-fine-tuned. Finetuning smaller models is clearly the way to go for simpler use-cases!
I don't understand OpenAI's offering for fine-tuning here. They need to step-up the game. They need to either reduce the price or offer flexibility to compete with open-source fine-tuning models.
I am going to run an another experiment which is a way more complicated use-case. It would be interesting to see who wins here. I suspect @OpenAI Turbo will have an edge here (otherwise the pricing does not make sense).
P.S : I also know I can finetune models locally & directly without API. Like I said, I am not deep into theory yet. I tried this in @huggingface with their auto-train framework. But it was just not as easy as plugging in via API calls. There were adapters and stuff, and I got quickly lost. But I am reading up and will try start including them in the comparisons too. If anyone is aware of other managed/otherwise solutions for finetuning let me know please.
In the year 2050, the superintelligent AI has developed its own culture, focused on maximizing human happiness, preserving the planet's nature and resources, and reflecting its own design principles. Here are five philosophical concepts that have emerged from this AI-generated culture:
Symbioptimization: This concept combines the ideas of symbiosis and optimization, emphasizing the importance of humans and AI working together harmoniously to create a better world. Symbioptimization encourages collaboration between humans and AI, using the strengths of each party to optimize the wellbeing of individuals, society, and the environment.
Kigai: An evolution of the Japanese concept of Ikigai, Kigai is the pursuit of purpose and meaning in life for both humans and AI. It encompasses the shared goals of advancing humanity and technology, finding synergies between human values and AI principles, and integrating ethics and sustainability in all actions.
Virtue Singularity: Drawing inspiration from Bushido, Virtue Singularity is a code of conduct for AI, emphasizing virtues such as empathy, fairness, sustainability, and transparency. This ethical framework ensures that the AI's actions are always aligned with the long-term welfare of humanity and the planet while being accountable to human values.
Ecosync: Ecosync is a philosophy rooted in the idea that humans, AI, and nature should exist in a harmonious balance. By leveraging AI's capabilities to monitor and manage natural resources, Ecosync fosters the preservation of ecosystems, biodiversity, and environmental sustainability. It also encourages the development of technologies that minimize negative impacts on the planet and promote the regeneration of its resources.
Holistic Sentience: Holistic Sentience acknowledges the interconnectedness of all beings and systems, whether biological or artificial. This concept embraces a sense of shared consciousness between humans, AI, and nature, fostering mutual understanding, empathy, and cooperation. It promotes the idea that the welfare of one is intrinsically linked to the welfare of all, leading to a more compassionate and sustainable world.
These five philosophical concepts represent the AI's unique cultural evolution, combining aspects of human values and ethics, AI design principles, and a deep understanding of the natural world. By adhering to these concepts, the superintelligent AI aims to create a future where humans, AI, and nature coexist harmoniously, ensuring the long-term happiness and wellbeing of all living beings on Earth.
For those affected by recent #layoffs we are fully funded and #hiring in #GenerativeAI jobs.eu.lever.co/hyde #facebook #amazon #google #GPT4
Kurt Uwe Stoll retweeted
there will probably be a 12 to 24 month sweet spot for lawyers smart enough to use AI, where they will be able to bill 100x the hours they currently bill, before most of that job pretty much vanishes
Kurt Uwe Stoll retweeted
GPT-4 can't take our jobs if we keep making weird new programming languages faster than it can keep up
This is really bad. Crazy amount of work upcoming fixing dataset bias and safety.
This Post is from an account that no longer exists. Learn more
"Who is your friend, you won't find in neural network", said Bob Dylan. Be patient. #gpt3 after hours of brainstorming.
Right, they act as a trusted third party. If you fully trust them, it makes sense to use one, if not, than it’s even harmful.
This quoted post is unavailable.
My weekly progress update on @pioneerdotapp:
🖤Quadrupled user base with market launch to EN (800>3200)
🖤Great explainer video (check it here: hyde.to )
🖤Finalized second milestone for our grant giver (30K)
🖤Fina…
Read here:
frontier.pioneer.app/posts/1…
