Today in AI

In partnership with

• News roundup

• Today’s big story

• Research spotlight

• This week on the podcast

News roundup

The top stories in AI today.

NEW LAUNCHES

The latest features, products & partnerships in AI

• Diaflow.io lets anyone create AI apps without the need for coding

• Microsoft, The Vatican partner on AI-powered digital replica of St. Peter’s Basilica

• UAE and Google join forces to transform digital landscape

• Washington Post launches AI chatbot to answer reader questions

• Instagram tests AI-generated profile pictures for users

• Bagel Explores: training-time and inference-time techniques to advance AI reasoning (Sponsored)

MONEY FLOWS

Economics, Wall Street, investing & fundraising

• Northflank raises $22.3M to simplify cloud infrastructure for devs

• Freshworks cuts 660 jobs in AI-focused restructuring

• AI startup funding soars to new heights in Los Angeles

• How AI startup Connecty aims to help enterprises maximize their data value

• Nvidia reclaims top spot as AI chip demand surges

• Amazon plans multibillion-dollar investment in AI firm Anthropic

AI MODELS

Deployment, research, training & infrastructure

• Generative AI models in healthcare require a reassessment of their reliability

• Stanford researchers probe LLMs for consistency and bias

• OpenAI’s Orion model is reportedly only somewhat better than GPT-4

• How one computer scientist’s stubbornness inadvertently sparked the deep learning boom

• DeepMind open sources its groundbreaking AlphaFold3 AI protein predictor

• AI video models try their best — but still struggle — to replicate real world physics

IMPLEMENTATION

Announcements, strategies & programs

• Businesses are increasing spend on AI in 2025 — here’s how they’ll spend it

• The top AI video generators by use case

• Alaska schools adopt flawed policy based on fake AI-generated studies

• The future of AI-driven retail

• How AI is changing our perception of risk and what it means for industry

• How software developers should be incorporating AI into their workflows

What’s happening in AI right now

Machine vision systems take on real-world challenges

The theoretical debates about AI are giving way to practical applications. The latest wave of machine vision systems is tackling tangible problems - from preserving ancient artifacts to identifying security threats.

Beyond simple recognition

At Ritsumeikan University, researchers have cracked a persistent challenge in computer vision: creating detailed 3D models from single 2D photographs. Their neural network reconstructed an ancient stone relief from Indonesia's Borobudur Temple using a 134-year-old photograph. The breakthrough lies in "soft-edge detection" - allowing AI to interpret subtle variations in texture and shadow that traditional systems miss.

The applications stretch beyond archaeology. Architects could analyze historical buildings, disaster response teams could assess structural damage, and engineers might gain new ways to inspect infrastructure without physical access.

Real-world tests

The Pentagon's new Bullfrog system demonstrates how machine vision pairs with physical systems for autonomous defense. This AI-enabled platform identifies and tracks small drones - a task that has long challenged human operators.

In Texas, Wytec International is putting similar principles to work in schools. Their integrated network of sensors and cameras has shown 90% accuracy in laboratory testing for threat detection. The system combines visual detection, acoustic analysis, and pattern recognition - showing how multiple AI models can work in concert.

The reality check

Laboratory success doesn't always translate smoothly to real-world deployment. Recent studies of AI vision and language models in healthcare reveal how minor variations in input can significantly impact accuracy. Physical-world AI faces similar challenges - varying lighting conditions, weather, and unexpected obstacles all pose potential complications.

What's next

The coming year will likely reveal which approaches to physical-world AI prove most effective. Early signs suggest success lies not in complete automation, but in systems that enhance human capabilities while maintaining clear boundaries. The organizations showing the most promise are those taking measured steps rather than rushing to automate everything at once.

The real test for machine vision won't just be technical performance - it will be finding the right balance between capability and reliability in real-world conditions. The technology is ready for deployment; the question now is where it makes the most sense to use it.

We publish daily research, playbooks, and deep industry data breakdowns. Learn More Here

The Bagel team just published new research on two complementary methods for advancing AI reasoning: training-time and inference-time techniques.

Training-Time Enhancements involve refining model structures. Methods like Parameters Efficient Fine-Tuning (PEFT) optimize learning efficiency by targeting specific neural pathways within fixed frameworks, while approaches like WizardMath’s 3-step reasoning improve structured reasoning.

Inference-Time Enhancements focus on on-the-fly thinking, using Chain of Thought prompting to activate logical processing without extra training. Techniques like Self-Consistency for validation and Program of Thought for coding tasks enable precise, stepwise reasoning.

The findings suggest combining both approaches: training-time builds foundational reasoning, while inference-time optimizes its application. Bagel Network supports this advancement through open-source infrastructure, empowering a community-driven AI evolution though monetizable open source AI.

About Bagel: Bagel is an AI & cryptography research lab, building the world's first monetization layer for open-source AI.

This week on the podcast

Can’t get enough of our newsletter? Check out our podcast Future-Proof.

In this episode, the hosts Anthony Batt and Shane Robinson with guest Joe Veroneau from Conveyor discuss outsmarting paperwork. Conveyor is a company that helps automate security reviews and document sharing between companies. They use AI technology, specifically language models, to automate the process of filling out security questionnaires. This saves customers a significant amount of time and improves the quality of their responses.