Hey there! If you missed any sessions from VB Transform 2023, don’t worry. You can register to access the on-demand library and catch up on all the amazing content. Just click here to register.
In an exciting announcement, OpenAI has introduced a new feature to its GPT-3.5 Turbo large language model (LLM). Users now have the ability to fine-tune the model with their own proprietary data, thanks to a partnership with Scale AI. This collaboration allows enterprises to train and deploy the model at scale, providing customized solutions for their specific needs.
Alexandr Wang, the CEO of Scale AI, expressed his enthusiasm about the partnership, stating that the combination of a powerful base model like GPT-3.5 and fine-tuning expertise from Scale AI can produce more accurate and efficient results. Wang compares the value of fine-tuning to the optimization process in software development, emphasizing its critical role in achieving the best possible outcomes.
Scale AI: Your Expert in Fine-Tuning
Scale AI brings together OpenAI’s advanced base model, GPT-3.5, with its own fine-tuning expertise and industry-leading Data Engine. This partnership aims to empower every company to create custom state-of-the-art models tailored to their specific business needs. Scale’s Data Engine accelerates model development by generating prompts and ranking model outputs.
The press release from Scale reveals that the company already specializes in fine-tuning various commercial and open-source models. As OpenAI’s preferred partner for GPT-3.5 fine-tuning, Scale AI is excited to leverage the powerful APIs provided by OpenAI to help enterprises build highly efficient custom LLMs while reducing costs.
Event
VB Transform 2023 On-Demand
Didn’t have time to attend sessions from VB Transform 2023? No worries! You can register now to access the on-demand library and catch up on all the amazing content.
Brad Lightcap, the COO of OpenAI, acknowledges the value that Scale AI brings in extending the power of fine-tuning to more companies. Leveraging Scale AI’s experience in enterprise AI, businesses can effectively apply OpenAI models to address their unique requirements.
Promising Results from a Fine-Tuning Pilot Program
Scale AI, a renowned unicorn company known for its data labeling services, has had an eventful year. Following some layoffs earlier in the year, the company made a comeback with the launch of its full-stack generative AI platform in May 2021, solidifying its position in the market.
The company highlights a case study involving its fine-tuning capabilities in collaboration with fintech company Brex. Brex had been using GPT-4 for generating high-quality expense memos to facilitate compliance requirements for employees. However, they wanted to explore if a fine-tuned GPT-3.5 model could improve cost efficiency and latency while maintaining quality.
According to Scale’s press release, they fine-tuned the GPT-3.5 model using annotated data from Brex, generated by Scale’s Data Engine. The results were impressive, with the fine-tuned model outperforming the stock GPT-3.5 turbo model 66% of the time.
Brex’s CEO, Henrique Dubugras, expressed his satisfaction with the partnership with OpenAI and Scale AI. He emphasized that fine-tuning GPT-3.5 has been a game changer for Brex, enabling them to deliver high-quality AI experiences comparable to GPT-4 at a much lower cost and latency. This advancement unlocks new possibilities and capabilities that were previously unattainable.
VentureBeat is the go-to platform for technical decision-makers seeking knowledge about transformative enterprise technology. Stay updated with our Briefings.
from GPT News Room https://ift.tt/AgMmwKW
No comments:
Post a Comment