Meta's New Human-like AI Model
Plus: Google’s human attention AI can enhance UX. OpenAI’s updateS on GPT-3.5 & 4.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 41st edition of The AI Edge newsletter. This edition brings you Meta’s new human-like AI model for image creation.
And a huge shoutout to all our readers out there. We appreciate you! 😊
🤖 Meta’s new human-like AI model for image creation
☁️ Google’s human attention models can enhance UX
🤗 OpenAI’s massive update on GPT-3.5 and GPT-4 APIs
📚 Knowledge Nugget: On Detecting Whether Text was Generated by a Human or an AI Language Model by
Let’s go!
Meta’s new human-like AI model for image creation
Meta has introduced a new model, Image Joint Embedding Predictive Architecture (I-JEPA), based on Meta’s Chief AI Scientist Yann LeCun’s vision to make AI systems learn and reason like animals and humans. It is a self-supervised computer vision model that learns to understand the world by predicting it.
The core idea: It learns by creating an internal model of the outside world and comparing abstract representations of images. It uses background knowledge about the world to fill in missing pieces of images, rather than looking only at nearby pixels like other generative AI models.
Key takeaways: The model
Captures patterns and structures through self-supervised learning from unlabeled data.
Predicts missing information at a high level of abstraction, avoiding generative model limitations.
Delivers strong performance on multiple computer vision tasks while also being computationally efficient. Less data, less time, and less compute.
Can be used for many different applications without needing extensive fine-tuning and is highly scalable.
Why does this matter?
As a first of its kind, this model represents a significant development in the field of AI, getting us a step closer to human-level intelligence in AI. It is intended to overcome key limitations of even the most advanced AI systems today.
Google’s human attention models can enhance UX
Google presents new research in the area of human attention modeling. It showcases how predictive models of human attention can enhance user experiences, such as image editing to minimize visual clutter, distraction or artifacts, and image compression for faster loading of webpages or apps.
Attention-guided image editing: Human attention models usually take an image as input and predict a heatmap as output. The heatmap is evaluated against ground-truth attention data, typically collected by an eye tracker or approximated via mouse hovering/clicking. Edits based on the heatmap can significantly change an observer’s attention to different image regions. For example, reducing clutter in the background in video conferencing may increase focus on the main speaker.
Why does this matter?
While this showcases how such models can enable delightful user experiences, it also guides ML models toward more intuitive human-like interpretation and model performance. And it encourages further research in this direction which could enable applications such as using radiologists’ attention on medical images to improve health screening or diagnosis or using human attention in complex driving scenarios to guide autonomous driving systems.
OpenAI’s massive update on GPT-3.5 & GPT-4 APIs
OpenAI announced exciting updates, including more steerable API models, function calling capabilities, longer context, and lower prices.
Function calling is now available to enable LLMs to work more effectively and efficiently interact with your programs/tools.
The latest GPT-3.5-Turbo model is available with support for 16K context (~20 pages of text)
25% cost reduction on input tokens for gpt-3.5-turbo
75% cost reduction on the embeddings model
The data privacy and security assurances implemented on March 1 remain consistent across all models. The user’s API data will not be utilized for training purposes
Why does this matter?
With their new update on more steerable API models, users can have more influence on the specific outcomes they desire, making the models even more versatile and adaptable to individual needs. The cost reductions make their services even more accessible.
Knowledge Nugget: On Detecting Whether Text was Generated by a Human or an AI Language Model
Machines can now produce text that closely resembles human-generated text. Isn’t it? But let me tell you, it can raise concerns about the potential for misinformation, spam, cheating, and impersonation.
Interesting? In this in-depth and thoughtful piece,
will explain the two working methods by researchers to detect machine-generated text, despite the increasing sophistication of large language models like GPT-3 and ChatGPT.In the two recent papers:
The first paper suggests a technique called "DetectGPT Zero-Shot Machine-Generated Text Detection using Probability Curvature," which aims to determine if a specific LLM generated a text passage.
The second paper proposes a "watermarking" method to embed unique identifiers during text generation, allowing users to check for the presence of a watermark.
Notably, both papers are authored by individuals from academia, highlighting the continued innovation in AI research from universities.
Why does this matter?
Detecting machine-generated text is significant due to the following:
It may help combat the spread of false information.
Identification of machine-generated text reduces spam and improves online interactions.
Detection methods prevent identity theft and manipulation through machine-generated impersonations.
Detection supports fair evaluation and prevents cheating in educational settings.
Detecting machine-generated text fosters transparency and ensures reliable information sources.
What Else Is Happening
🎨 Adobe introduces Generative Recolor- Transform your palette with text prompts! (Link)
🚀 AMD introduces Instinct MI300X, world's most advanced accelerator for generative AI (Link)
💰 Accenture to invest a whopping $3 billion in AI over the next three years (Link)
🤖 NVIDIA’s ATT3D framework simplifies text-to-3D modeling (Link)
🤝 Hugging Face collaborates with AMD to ensure native optimization of its models (Link)
Trending Tools
Framer AI: Design with ease using AI. Choose from infinite color palettes and typeface combos. Customize with AI-generated copy and color shuffling.
Playbook AI: Visual cloud storage for AI art. Organize your prompts, art, and iterations so you can focus on your craft.
Greenifs.Ai: Detects greenwashing errors on social media. Ensures compliance with green marketing guidelines and safeguards your brand’s reputation.
Uizard Autodesigner: Generate multi-screen designs with a simple text prompt using AI. Bring your vision to life in seconds
Trickle 1.0: AI-driven workspace blending notes, tasks, and knowledge base for you and your team. Get inspired, write faster, and work smarter with GPT-4.
Scalenut: Automates SEO workflow. Saves time and delivers organic traffic. Provides powerful SEO strategy, creates content at scale, and decodes search guidelines.
Superpowered: AI copilot for meetings. No bots or recordings, just really good AI notes.
Favikon: First AI-based platform ranking creators on social media for actionable insights in social media strategy.
That's all for now!
If you are new to ‘The AI Edge’ newsletter. Subscribe to receive the ‘Ultimate AI tools and ChatGPT Prompt guide’ specifically designed for Engineering Leaders and AI enthusiasts.
Thanks for reading, and see you tomorrow.