🖼️ AI understands IMAGES?

Hey TEDchats Fam!

Happy Wednesday! Remember the childhood joy of "connect-the-dots" puzzles? As adults, the dots have evolved, and so has the game. Now, AI connects the dots between text and images, making way for groundbreaking tech advancement.

👀 SNEAK PEEK

🚨FEATURE STORY

What Does GPT-4 with Vision Mean for Us?

Have you ever wondered what it'd be like if text-generating AI could also understand and interpret images? OpenAI’s GPT-4 provides just that, but not without its share of controversies and concerns.

GPT-4 Was Never Just About Text…

When OpenAI first unveiled GPT-4, it wasn't just about text. The announcement brought to light the model's multimodality: its impressive ability to make sense of both images and text. For instance, GPT-4 could identify a Lightning Cable adapter simply by viewing a picture of a plugged-in iPhone. However, this groundbreaking innovation was clouded by OpenAI's decision to withhold the model's image features from the public due to speculated concerns of misuse.

The Dark Side of AI Having Vision

The mystery behind OpenAI's reservations about GPT-4's image-analyzing capabilities was finally unveiled. A technical paper recently published by the company brought forward not just the brilliant features of GPT-4V(ision), an extension of GPT-4, but also its potential pitfalls. In this paper, OpenAI outlined various concerns that were internally investigated, including the evaluation of harmful content, harms of representation, privacy, cybersecurity and multimodal jailbreaks — all being important concerns (read more). For example, how can it be ensured that GPT-4V won't be used maliciously to break CAPTCHAs or provide methods for producing dangerous drug substances solely from an image?

OpenAI's Response to Challenges

OpenAI has since been proactive, working diligently to mitigate potential risks. They have implemented safeguards to prevent certain misuse or biases, like identifying individuals based on age or race. They have additionally improved their transparency about GPT-4V by releasing research updates and internal experiment data. However, the paper makes it clear: like all AI, GPT-4V has its limits, including its proneness to ‘hallucinate’ or misinterpret obvious objects, and it's essential to be aware of them.

🍕 QUICK BITES

Amazon's Billion-Dollar AI Play

Amazon takes a strategic leap into the AI frontier, agreeing to invest an initial $1.25 billion for a minority stake in Anthropic, with options to invest up to $4 billion. This strategic move aligns them with top AI innovators, aiming to enhance customer experiences and advance AI technology.

Voice Revolution on Spotify

Soon, podcasters could instantly translate and reproduce their episodes in another language, all while retaining their unique voice. Spotify's AI-driven feature, powered by OpenAI, will be first tested on a select group of podcasters, but is set to transform the podcasting landscape.

Snapchat & Microsoft Team Up

Snapchat's 'My AI' chatbot will now include Sponsored Links in partnership with Microsoft, albeit still in an experimental phase. If you're discussing dinner in the chat, for instance, you might see a link to a nearby restaurant. It's a novel approach, embedding ads directly within AI-driven conversations.

🧰 AI TOOLBOX 

🤯 MIND SPARK

AI Uses X-Ray Data to Predict Lung Disease Mortality: Researchers developed an AI model to analyze chest X-ray images to predict the likelihood of developing lung diseases, such as asthma and lung cancer, as well as to predict the associated risk of death (read more).

What is Happening With Robotaxis? Driverless taxi services, known as robotaxis, seem to be a current hype among the growing trend of autonomous vehicles. Would you ride in a driverless taxi? (read more)

J.P. Morgan Evaluates AI’s Influence on the Market: In a recent report, J.P. Morgan & Co. evaluate the potential effect of AI on the economy, including the trading and job market (read more).

🎨 AI ART