New AI Model Could Make Siri Way Faster and Smarter 🔥

Welcome to another edition of Horizon AI,

Apple has introduced a new model that could transform the way people have interacted with Siri so far. They even claim that this model surpasses GPT-4 in certain aspects.

Let's jump into it!

Read Time: 3.0 min

Here's what's new today in the Horizon AI

• A Chart to Kickstart Your Day 📈: Visualizing Nvidia’s Revenue, by Product Line (2019-2024)

• Apple's new AI model ReALM 'Surpasses GPT-4' 🔥

• AI Research: Anthropic Researchers Expose "Many-Shot Jailbreaking" Vulnerability in LLMs

• 🚨 Job Opportunity

• AI Tutorial: Create any design for free with Microsoft Designer

• AI Tools to check out

• The Latest in AI and Tech 💡

A chart to kickstart your day

Visualizing Nvidia’s Revenue, by Product Line (2019-2024)

• Today, data center processors for analytics and AI have quickly become Nvidia’s largest source of revenue, surpassing the company’s historical main business line of GPUs

• This trend is expected to continue as the demand for AI chips keeps growing.

Source

AI News

APPLE

Apple's new AI model ReALM 'Surpasses GPT-4' 🔥

In a research paper, Apple unveiled ReALM (Reference Resolution As Language Modeling), a novel AI method for more efficiently understanding contextual information that could supercharge virtual assistants like Siri.

Details:

• This AI model can understand ambiguous references to on-screen entities as well as conversational and background context, enabling more natural interactions with voice assistants

• It uses a new approach of converting any contextual data, such as images, websites, and apps, into text, which is easier to parse by a Large Language Model. This allows it to bypass the complex image recognition capabilities needed by models like GPT-4.

• The smallest ReALM models performed like GPT-4 but with fewer parameters, making them ideal for on-device use. Adding more parameters to ReALM significantly boosted its performance over GPT-4.

For instance, while browsing a website, if you want to call the business, just saying 'call the business' would prompt Siri to interpret your request based on context. It can recognize a labeled phone number on the page and initiate the call without additional input from you.

Source

AI RESEARCH

Anthropic Researchers Expose "Many-Shot Jailbreaking" Vulnerability in LLMs

Researchers from AI startup Anthropic have published findings detailing a novel technique that can coerce large language models (LLMs) into providing harmful or unethical responses that they are typically programmed to avoid.

The method, dubbed "many-shot jailbreaking," exploits LLMs' in-context learning abilities that allow them to understand and apply new instructions within a given prompt.

Details on the many-shot jailbreaking technique:

• Involves inserting a series of fabricated dialogues into the prompt

• These carefully crafted dialog sequences can manipulate LLMs to violate their training

• The likelihood of eliciting a harmful response increases with the number of dialogues

The researchers claim they are sharing their findings in hopes of accelerating the development of defensive strategies against such vulnerabilities.

Source

JOB OPPORTUNITY

🚨 Job Opportunity in the US 🚨

Hello everyone, Gina here, I’ve some exciting news to share!

I've worked as a Data Analyst and Data Engineer (you can check out my Linkedin profile here), and in my day-to-day work, I see plenty of Data and AI job opportunities at some fascinating companies.

I’ve been connecting candidates to companies in a somewhat ad-hoc way recently, but I want to start trialing this on a larger scale.

So, if you are looking for your next gig, click on the link below to help us match you with jobs!

👉 Form to fill out here

Employers - don't feel left out! If you are hiring for Data and AI talent or know someone who is, then reply to this email.

Please note that this opportunity is specifically for individuals residing in the United States.

AI Tutorial

Create any design for free with Microsoft Designer

With Microsoft Designer you can leverage AI and Dall-E 3 to create visuals for your social media and more

1. Go to Microsoft Designer and log in

2. Go to the bottom of the home page and select a design type

3. Create the basic design

Select the "Ideas" icon at top right

Enter a prompt, for example: 'an Instagram post wishing a happy Mother’s Day.'

Designer AI will give you several suggestions to choose from; just select the one you like the most.

4. Add AI generated images

You can create and add images generated with DALL-E 3. Just click on 'Visuals,' then 'Generate,' and write a prompt.

Microsoft Designer will generate 4 images. You can put them wherever you want in your design or use them as a background.

5. AI Suggestions

Click on the "Ideas" icon at the top again, this time it will give you design suggestions based on what you created so far

You can select one and add the finishing touches in a few minutes (change colors, fonts, etc)

6. Download

Click on ‘Download’ and choose the file type

AI Tools to check out

📽 Katalist: Create visual stories with consistent characters and scenes using generative AI.

🎧 iListen: Summarise any article or webpage into concise, easy-to-digest podcasts.

✅ Airtrain: Query and compare a large selection of open-source and proprietary models at once.

📊 Flowsage: Effortlessly design, collaborate, and optimize flowcharts with the power of AI

✨ SynthesysX: Create your own versions of any image you find online.

The latest in AI and Tech

AWS and Mistral AI expand collaboration

Amazon customers can now access Mistral Large via Amazon Bedrock. Additionally, the European AI startup will leverage AWS hardware, including AWS Trainium and AWS Inferentia, to train and deploy future foundation models on Amazon Bedrock.

Artists sign open letter warning against 'predatory' use of AI in music

In an open letter by the Artist Rights Alliance, over 200 prominent artists, songwriters, and producers, including Billie Eilish, Pearl Jam, and Nicki Minaj, have voiced their concerns over irresponsible AI training practices within the music industry. The letter characterizes such practices as an "assault on human creativity" that "must be stopped."

Yahoo acquires AI news platform Artifact

Yahoo announced on Tuesday its acquisition of Artifact, an AI-driven news platform co-founded by the creators of Instagram. Through this agreement, Yahoo plans to integrate Artifact's AI technology across its news and online services throughout the United States.

FDA greenlights Prenosis' AI-powered Sepsis diagnostic tool

Prenosis announced its AI-powered sepsis diagnostic tool as the first FDA-approved one. Prenosis' tool, called Sepsis ImmunoScore, assists clinicians in assessing a patient's risk of sepsis and generates an overall risk score to aid in the diagnosis.

OpenAI removes Sam Altman's ownership of its Startup Fund

According to a filing with the U.S. Securities and Exchange Commission (SEC), OpenAI has revised the governance structure of its venture capital fund focused on backing AI startups. The restructuring has effectively removed OpenAI's CEO, Sam Altman, from owning or controlling the fund.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻