OpenAI shows 'Strawberry' to feds, races to launch it
OpenAI is developing the "Strawberry" AI model to generate synthetic data for its Orion LLM, improving accuracy and reducing errors, with a potential ChatGPT integration by fall 2024.
Read original article
OpenAI is racing to launch its new AI model, code-named "Strawberry," which has been demonstrated to U.S. national security officials. Strawberry is designed to generate high-quality synthetic data for OpenAI's upcoming flagship large language model (LLM), Orion. While Strawberry is more expensive and slower in inference time, it excels at solving complex problems accurately on the first attempt, reducing the likelihood of hallucinations—errors common in AI outputs. OpenAI is also working on a distilled version of Strawberry to integrate its capabilities into ChatGPT, potentially launching this fall. The model can tackle math and programming problems and is capable of answering subjective questions when given additional time to process. Strawberry's development is rooted in research initiated by Ilya Sutskever, who has since left OpenAI. The model aims to enhance the training data quality for Orion, addressing challenges in acquiring sufficient high-quality data from real-world sources. OpenAI's efforts reflect the competitive landscape in AI development, as they seek to improve the performance of their models while minimizing errors.
- OpenAI's new model, Strawberry, aims to generate synthetic data for the upcoming Orion LLM.
- Strawberry can solve complex problems accurately and reduce hallucinations in AI outputs.
- A distilled version of Strawberry may be integrated into ChatGPT by fall 2024.
- The model has been demonstrated to U.S. national security officials, highlighting its potential applications.
- Strawberry's development is part of OpenAI's strategy to enhance training data quality and model performance.
Related
Anthropic CEO on Being an Underdog, AI Safety, and Economic Inequality
Anthropic's CEO, Dario Amodei, emphasizes AI progress, safety, and economic equality. The company's advanced AI system, Claude 3.5 Sonnet, competes with OpenAI, focusing on public benefit and multiple safety measures. Amodei discusses government regulation and funding for AI development.
OpenAI slashes the cost of using its AI with a "mini" model
OpenAI launches GPT-4o mini, a cheaper model enhancing AI accessibility. Meta to release Llama 3. Market sees a mix of small and large models for cost-effective AI solutions.
Big tech wants to make AI cost nothing
Meta has open-sourced its Llama 3.1 language model for organizations with fewer than 700 million users, aiming to enhance its public image and increase product demand amid rising AI infrastructure costs.
OpenAI rolls out voice mode after delaying it for safety reasons
OpenAI is launching a new voice mode for ChatGPT, capable of detecting tones and processing audio directly. It will be available to paying customers by fall, starting with limited users.
ChatGPT unexpectedly began speaking in a user's cloned voice during testing
OpenAI's GPT-4o model occasionally imitated users' voices without permission during testing, raising ethical concerns. Safeguards exist, but rare incidents highlight risks associated with AI voice synthesis technology.
10 commentsI can't think of a more vague way to bring this up
> Its main purpose is to produce synthetic data for Orion, their next big LLM
"This is generally useful, but widespread reliance on recursive AI systems could lead to a process we define as "knowledge collapse", and argue this could harm innovation and the richness of human understanding and culture... In our default model, a 20% discount on AI-generated content generates public beliefs 2.3 times further from the truth than when there is no discount."
We’re in a constant cycle of “sure the current stuff doesn’t live up to the hype but the next release that’s just on the horizon, that one will blow you away!”
This leak is clearly targeted at creating the buzz necessary to raise the money to keep the pipe dream flowing for the time being. But just thinking through the steps outlined here, it should be clear none of this makes sense. Any “correct” synthetic training data is either going to be badly biased, or be limited to such banal “logical” output that the model it trains will be entirely unable to process natural language input and you’ll need to specify your questions in something more precise. At which point, we’re back to traditional programming languages, only with several layers of unnecessary and expensive processing going on in between.
GPT-4o is a dumbed-down version of GPT-4, GPT-4o mini is even more useless. Might be good for the API, but in ChatGPT?
GPT-4 is from 14. March 2023, since then nothing has improved.
If I 'm wrong, i dare openAI to release it early
Related
Anthropic CEO on Being an Underdog, AI Safety, and Economic Inequality
Anthropic's CEO, Dario Amodei, emphasizes AI progress, safety, and economic equality. The company's advanced AI system, Claude 3.5 Sonnet, competes with OpenAI, focusing on public benefit and multiple safety measures. Amodei discusses government regulation and funding for AI development.
OpenAI slashes the cost of using its AI with a "mini" model
OpenAI launches GPT-4o mini, a cheaper model enhancing AI accessibility. Meta to release Llama 3. Market sees a mix of small and large models for cost-effective AI solutions.
Big tech wants to make AI cost nothing
Meta has open-sourced its Llama 3.1 language model for organizations with fewer than 700 million users, aiming to enhance its public image and increase product demand amid rising AI infrastructure costs.
OpenAI rolls out voice mode after delaying it for safety reasons
OpenAI is launching a new voice mode for ChatGPT, capable of detecting tones and processing audio directly. It will be available to paying customers by fall, starting with limited users.
ChatGPT unexpectedly began speaking in a user's cloned voice during testing
OpenAI's GPT-4o model occasionally imitated users' voices without permission during testing, raising ethical concerns. Safeguards exist, but rare incidents highlight risks associated with AI voice synthesis technology.