OpenAI announce GPT fine-tuning

Today’s Edition

Hey there! Welcome to today's edition, packed with all the hottest news from the world of AI.

• 📖 Today, we examine OpenAI's recent announcement that developers can now customise GPT-3.5 Turbo with their own data to suit their specific use cases. This development is a positive step towards democratising AI for businesses of all sizes.

• 🖥️ Today’s product section highlights Canvas, a potentially indispensable tool for all startups, alongside Dili, which is specifically designed for Private Equity and Venture Capital deal teams to enable more data-driven investment decisions.

• 👩🏼‍💻 Embrace the role play In sales, dealing with objections is one of the biggest challenges a salesperson can face. By feeding ChatGPT details about your product and your business, and a scenario you wish to rehearse, you can practice how to handle some of the common objections and devise strategies to overcome them.

• 🎤 Most valuable media features Midjourney's innovative Inpainting feature, illustrating one specific application for it, while also presenting a discourse between George Hotz and Eliezer Yudkowsky as they delineate their views on AI safety, acceleration, and associated subjects.

So grab a cup of coffee, sit back, and let's dive into the fascinating world of AI together. Enjoy!

Jenko

OpenAI announce GPT fine-tuning

Developers can now bring their own data to customise GPT-3.5 Turbo for their use cases.

Yesterday’s announcement from OpenAI on the availability of Fine-tuning for GPT-3.5 Turbo, with the upcoming addition of fine-tuning for GPT-4 slated for this fall definitely peaked my interest. This newly unveiled feature empowers developers to modify models so they can perform optimally for specific use cases, and it also enables the execution of these custom models on a large scale.

The significance of this update is underscored by early tests demonstrating that a fine-tuned GPT-3.5 Turbo can equal or even surpass the capabilities of base GPT-4 for particular specialised tasks. In line with OpenAI's commitment to customer data sovereignty, all information processed through the fine-tuning API remains under the exclusive ownership of the customer and will not be utilised by OpenAI or any other entity to train subsequent models.

Fine-tuning use cases

Since the release of GPT-3.5 Turbo, developers and businesses have been asking for the ability to customise the model to create unique and differentiated experiences for their users. With this launch, developers can now run supervised fine-tuning to make this model perform better for their use cases.

In the private beta, fine-tuning customers have been able to meaningfully improve model performance across common use cases, such as:

• Improved steerability: Fine-tuning allows businesses to make the model follow instructions better, such as making outputs terse or always responding in a given language. For instance, developers can use fine-tuning to ensure that the model always responds in German when prompted to use that language.

• Reliable output formatting: Fine-tuning improves the model's ability to consistently format responses—a crucial aspect for applications demanding a specific response format, such as code completion or composing API calls. A developer can use fine-tuning to more reliably convert user prompts into high-quality JSON snippets that can be used with their own systems.

• Custom tone: Fine-tuning is a great way to hone the qualitative feel of the model output, such as its tone, so it better fits the voice of businesses’ brands. A business with a recognisable brand voice can use fine-tuning for the model to be more consistent with their tone - think ad copy or internal communications at least partially written by AI.

In addition to increased performance, fine-tuning also enables businesses to shorten their prompts while ensuring similar performance. Fine-tuning with GPT-3.5-Turbo can also handle 4k tokens—double our previous fine-tuned models. Early testers have reduced prompt size by up to 90% by fine-tuning instructions into the model itself, speeding up each API call and cutting costs.

Companies integrating AI within their products and operational flows stand to gain significantly from the precision offered by fine-tuning ChatGPT to concentrate on particular functions, like code completion or sustaining a uniform tone. This enhancement transforms ChatGPT into a more streamlined tool, curtails development expenses, and facilitates businesses, even those that were previously restricted by budget, to incorporate this advanced technology into their offerings, even if operating with a limited development team.

This feels like a step in the right direction towards democratising AI for businesses of all sizes.

Products

Product of the Day

New Products

• 365 - Log meals with your phone camera. Save hours every month and let our AI Nutritionist count your calories for you.

• Endorsed - Screen job applicants faster and smarter with AI-written summaries, organised un-biased in order of best fit for the job, so you can advance top applicants faster.

• Musicfy - Meet Your AI Music Assistant Use AI to create music with your voice

• dili - Dili is built for Private Equity and Venture Capital deal teams to make more data-driven investment decisions.

• Kraftful - Copilot for product builders. Listen to millions of users in minutes with AI.

Our Agency

Learn

A day in the life of AI

Every day, we will showcase a workflow or task designed to help you save time in your job or personal life. Becoming skilled at promoting can be a superpower, so the goal of this section is to help you improve in this area for both personal and professional growth.

Personal or Professional: Professional
Workflow/Task: Sales Roleplay
Task time before AI: Impossible
Task Time with AI: <5 minutes

Embrace the role play in sales, dealing with objections is one of the biggest challenges a salesperson can face.

By feeding ChatGPT details about your product and your business, and a scenario you wish to rehearse, you can practice how to handle some of the common objections and devise strategies to overcome them. The prompt:

Most Valuable Media (MVM)

Most Valuable Image (MVI)

Picture in a picture? Let's take a look at using an image prompt with Midjourney's new inpainting feature!

First, I need a starting image. For this example, I have chosen a woman wearing a plain white t-shirt. Next, I need my reference image's URL. (See MJ's guide to image prompts for help with this.)

From my starting image, I use "Vary (Region)" and select a portion of the t-shirt. Then (with Remix Mode on) I add the URL first, and then a word prompt. In this case, I used the original prompt that produced the image of the flower illustration. (If you are using a non-MJ image, you will need to write a new word prompt.)

Ta-da! Now her shirt has a pretty graphic on it.

Most Valuable Tweet (MVT)

Most Valuable Video or Podcast (MVVP)