ChatGPT adds new feature

Reading time: 7 minutes  

 Newsletter | News | Tools 

Today we will discuss-

• 🤩OpenAI makes GPT-3.5 more attractive for developers 

• 👨‍🌾Meta's new AI model knows 100 languages

• 👨‍🎨AI chip market to be worth $67B in 2024

• 🕵️‍♂️9 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

OpenAI helps developers by introducing ‘fine-tuning’ feature to GPT-3.5 Turbo

🧐Context of the news - Language models such as ChatGPT are trained on a massive dataset of text, which means that they are able to learn general patterns about language. However, they may not be able to perform well on a specific task.

This is exactly where ‘fine-tuning’ can help. Fine-tuning is like taking a machine learning model that is already good at something and making it even better at that thing.

🤖News - OpenAI has introduced 'fine-tuning' feature to the GPT-3.5 Turbo, a model already fine-tuned for chat applications. The fine-tuned versions of GPT-3.5 Turbo can match or even outperform the base capabilities of GPT-4.

Now, developers can train GPT-3.5 Turbo on a dataset that is specific to their needs. This can be used to improve the accuracy of the model and to build in specific behaviors. As per the company, this feature wouldn’t be free of cost.

Here’s an example - Let's say you are a developer who wants to use GPT-3.5 Turbo to generate text for marketing copy. You could bring your own dataset of marketing copy to the model and train it on that. This would help the model to learn the specific style and tone of your marketing copy, and it would produce the content in your desired way.

The fine-tuning support for GPT-4 will arrive sometime later this fall.

Meta's new AI model can translate speech in nearly 100 languages

👩‍💻News - Meta has released an AI model called SeamlessM4T that can translate and transcribe close to 100 languages across text and speech. The model is available in open source.

This project by Meta is a significant breakthrough in the field of AI-powered speech-to-speech and speech-to-text. It is among the more ambitious efforts to date to combine translation and transcription capabilities into a single model.

👩‍🎓What is its specialty?

• SeamlessM4T does not need a separate language identification model to identify the source language. This makes it more efficient and accurate than previous models.

• Meta developed the model by collecting publicly available text and speech data from the web. The company claims that the data was not copyrighted and came primarily from open source or licensed sources.

• The company asserts that SeamlessM4T does not generate an excessive amount of toxic text in its translations, which is a common issue with translation and generative AI text models.

• Using the model for long-form translation and certified translations has been inadvisable by the company.

AI chip market to grow rapidly and reach $119 billion by 2027

🧑‍🌾News - Gartner, a market analyst, has predicted that the market for AI chips will grow at an annual rate of more than 20%. The market is expected to be worth $53.4 billion in 2023, up 20.9% from 2022.

In 2024, the market is expected to grow by 25.6% to $67.1 billion. By 2027, AI chips revenue can double the size of the market in 2023, reaching $119.4 billion.

🧑‍🎨What’s more? As the use of AI-based workloads in the enterprise matures, many more industries and IT organizations will deploy systems that include AI chips.

In the consumer electronics market, Gartner analysts estimate that the value of AI-enabled application processors used in devices will reach $1.2 billion by the end of 2023, up from $558 million in 2022.

As per the analysts, many organizations will use custom AI chips instead of discrete GPUs for AI workloads, especially generative AI.

👩🏼‍💻What else is happening?

• Study shows potential for generative AI to increase access and efficiency in healthcare

• Amazon's 'Thursday Night Football' broadcast will use AI to tell fans if teams should go for it on fourth down

• Meta confirms AI ‘off-switch’ incoming to Facebook, Instagram in Europe

• Zadie Smith, Stephen King and Rachel Cusk’s pirated works used to train AI

• More teachers embracing Artificial Intelligence in classrooms, says report

🎮AI-based games

• Left 4 Dead - It is a first-person shooter set in a post-apocalyptic world, where a pandemic has turned most of the population into mutated creatures. As part of a survivor team, you must navigate hazardous environments, complete objectives, and reach safety while facing waves of infected enemies.

• Black & White - It is a simulation game set in a fantasy world where you take on the role of a deity with the power to shape the land and influence the lives of its inhabitants. The game uses AI to give personality and autonomy to your followers and other creatures in the world.

• The SIMS series - A life simulation game that allows players to create and control virtual people, known as "Sims," and guide them through various aspects of their lives. AI manages various systems within the game, such as the progression of time, and the needs of the Sims.

👩🏼‍🚒Discover mind-blowing AI tools

1. AISuitUP - An AI-powered tool that generates professional headshots

2. Tinyquiz - A tool that allows teachers to generate tests and quizzes quickly and easily

3. Onri - An AI-powered people search tool that helps teams find experts within their organization

4. Motionshift - An AI-powered video creation tool that helps users create high-converting videos and ads quickly

5. Smalltalks - A platform that combines AI and human expertise to help people improve their online dating experience

6. Westidol - Allows users to create professional photoshoots using just a single selfie

7. Syllaby - A content strategy tool that helps businesses create a content plan for social media platforms

8. Impulse AI - An AI-powered tool that helps businesses generate marketing copy quickly and efficiently

9. Kaleido - A product management tool that helps development teams collect and organize user feedback

🥳Historic moments

• 1956: Mathematicians Allen Newell and Herbert A. Simon introduced the Logic Theorist, the first AI program that could prove mathematical theorems. Their breakthrough marked AI's nascent ability to reason and problem-solve, shaping AI's future path.

• 2006 - At the University of Toronto, renowned computer scientist Geoffrey Hinton led a demonstration highlighting deep neural networks' prowess, igniting AI's renaissance by inspiring extensive research and deep learning applications.

• 2021: Sam Altman unveiled CLIP, a pioneering multimodal AI which bridged language and images. The development marked an important step forward in creating AI systems with a deeper understanding of the world and the ability to interact with it in a more human-like manner.