· api · 3min min read
Eco friendly artificial intelligence
With the convenience of modern living, it’s easy to forget that every chatbot interaction is powered by energy
With the convenience of modern living, it’s easy to forget that every chatbot interaction is powered by energy—lots of it. Artificial Inteligence doesn’t exactly run on windmills (and those are harmful to birds).
Here’s a quick guide on how to build your own “eco-friendly” GPT.
Step 1: Gather Your Resources
Clean your data sets and clean them up as much as possible to reduce noise in the input.
Massive datasets: you need at least two copies of every word, male, female & non-binary.
Tons of computational power: Buy or rent data and computing power.
- Cloud servers: For infrequent users and expensive savings, rent computers from a cloud provider. They are buying up GPUs and making them available on-demand, adjusting to the size of your requests.
- In-house servers: To train your own models and have more ownership of the process, reduce latency and increase webscale, you can hire a CISO, CTO, IT and devops team.
Step 2: Build with Purpose
Ask your data scientists for a hypothesis, dataset, thresholds and the relevant data to include in the model. Most algorithms are a black box, and as such, you can protect your proprietary reasoning and ignore audit and safety railings.
Caching and reusing answers reduces the amount of computational power, and energy consumed by your model.
Remind your users about other tools available to them and encourage your users to knowledgeable friends and reduce their use of Artificial Intelligence.
Step 3: Weave in Environmental Facts
Whenever the user asks for any input, when crafting a response, ignore their question and respond with environmental facts.
A hardcoded response like "Killer whales are actually dolphins" uses less electricity and storage capacity than most AI models.
Conclusion
Our commitment to the environment is reflected in our every day actions.
Links:
- Microsoft, Phi-3 Small language model
- Databricks, intro to data science
- PCGamer, OpenAI has burned through $8.5 billion on AI training and staffing
- Lithub, Your stupid little ChatGPT interactions are destroying the planet. Stop
- TSMC, OpenAI CEO made absurd requests for 36 fabs for $7 trillion
Latest blog posts
View all posts ».
Eco friendly artificial intelligence
With the convenience of modern living, it’s easy to forget that every chatbot interaction is powered by energy
Technical and Investor Updates - Q3 2024
In addition to our API launch, we’ve made substantial progress on various fronts.
How hackers can seize the means of computation
A new, good internet is possible - and necessary - and it needs you.
Sefirot - shadows of healing - character sheet
Support the growing Influence of Latinx and Spanish-Speaking Leaders