Figure drops OpenAI in favor of in-house models

Figure AI, a company striving to introduce general-purpose humanoid robots for both commercial and residential applications, announced on X that it is ending its collaboration with OpenAI in favor of its own in-house AI development. The Bay Area-based robotics firm cited a “major breakthrough” in AI as the primary reason for this shift. While details remain scarce, founder and CEO Brett Adcock assured Techfulnews that Figure AI will unveil a groundbreaking humanoid AI advancement within the next 30 days.

Strategic Shift Away from OpenAI

OpenAI has played a notable role in Figure AI’s journey, previously collaborating on next-generation AI models for humanoid robots. In 2023, Figure AI secured $675 million in funding, bringing its valuation to $2.6 billion. To date, the company has raised $1.5 billion from investors, highlighting the immense financial backing behind its ambitious projects.

The decision to sever ties with OpenAI is unexpected, given the latter’s dominant position in artificial intelligence. Simply being associated with OpenAI typically offers companies a significant visibility boost. Just last August, Figure AI confirmed that its Figure 02 humanoid would utilize OpenAI’s models for natural language interactions.

The Challenge of Integration

Despite OpenAI’s expertise in AI-driven applications, Adcock pointed out that the integration process posed serious challenges. OpenAI operates on a massive scale, focusing primarily on general AI applications rather than embodied AI, which integrates artificial intelligence into physical systems such as robots. According to Adcock, vertical integration is the key to successfully scaling humanoid robotics.

“To solve embodied AI at scale in the real world, you have to vertically integrate robot AI,” Adcock told Techfullnews. “We can’t outsource AI for the same reason we can’t outsource our hardware.”

OpenAI’s Expanding Interest in Humanoid Robotics

Interestingly, OpenAI has been diversifying its investments in humanoid technology. The company is a major backer of 1X, a Norwegian robotics startup that focuses on humanoids for home applications. While many humanoid robot firms target industrial and warehouse automation, 1X has pivoted toward residential use cases.

Furthermore, OpenAI’s latest trademark application, filed with the U.S. Patent and Trademark Office (USPTO), suggests an interest in humanoid robotics. The filing references “user-programmable humanoid robots” and “humanoid robots with communication and learning capabilities”, fueling speculation that OpenAI might be developing its own hardware.

Figure AI’s Focus: Industry Over Homes

Although Figure AI has explored residential applications for its robots, its primary focus remains industrial deployments. This strategy aligns with automakers’ greater financial resources for testing and implementing emerging technologies. BMW, for example, began deploying Figure robots at its South Carolina factory in 2023, marking a significant milestone for the company.

The Competitive Landscape: Proprietary AI vs. Partnerships

The robotics industry remains divided on whether to develop proprietary AI models or rely on external partnerships. While Figure AI is doubling down on in-house development, other leading humanoid firms continue collaborating with AI specialists.

For instance, Boston Dynamics has partnered with the Toyota Research Institute to enhance its Atlas humanoid robot with advanced AI capabilities. The approach mirrors Apple’s vertically integrated ecosystem, where software is optimized specifically for proprietary hardware. However, achieving seamless hardware-software synergy is both complex and resource-intensive.

The Road Ahead for Figure AI

With its deep financial backing, Figure AI has been able to rapidly scale its hardware and software teams, recently relocating to a larger Bay Area office to accommodate its growth. The company’s shift away from OpenAI marks a pivotal moment in its mission to redefine humanoid robotics through proprietary AI models.

The next 30 days will be crucial, as Figure AI promises to showcase a technological advancement that has never been seen on a humanoid robot. If successful, this move could position the company as a leader in the humanoid robotics space, setting a new standard for vertically integrated AI-powered robots.

Google Unveils Gemma 3: A Leap Forward in Open AI Models for Developers

Sony’s Revolutionary RGB LED TV Technology: A Game-Changer in Display Innovation

The Evolution of Jet Bombers: From the B-45 Tornado to the B-21 Raider

How to Run Windows from a USB Drive: A Comprehensive Guide

News

What is AI Distillation and how Deepseek used it to blindside OpenAI

by George Mensah 2025-02-22

Artificial Intelligence (AI) is no longer just a buzzword—it’s a transformative force reshaping industries, from healthcare to finance. But as AI models grow more complex, a new challenge has emerged: how to make these powerful systems faster, smaller, and more efficient without sacrificing performance. Enter AI Distillation, a cutting-edge technique that’s revolutionizing the field.

In this article, we’ll explore what AI Distillation is, how it works, and how DeepSeek, a rising star in the AI space, used this innovative approach to outmaneuver industry giant OpenAI. Buckle up as we dive into the fascinating world of AI optimization and uncover the secrets behind DeepSeek’s success.

What is AI Distillation?

The Problem with Large AI Models

Modern AI models, like OpenAI’s GPT-4, are incredibly powerful but come with a catch: they’re massive. These models often have billions of parameters, requiring enormous computational resources and energy to train and run. This makes them expensive, slow, and inaccessible for many applications, especially on devices with limited processing power, like smartphones or IoT devices.

The Solution: AI Distillation

AI Distillation is a process that simplifies large AI models by transferring their knowledge to smaller, more efficient models. Think of it as teaching a student to summarize a textbook—instead of memorizing every detail, the student learns the key concepts and can apply them effectively.

In technical terms, AI Distillation involves training a smaller “student” model to mimic the behavior of a larger “teacher” model. The student model learns to replicate the teacher’s outputs while using fewer resources, making it faster and more cost-effective.

How Does AI Distillation Work?

Step 1: Training the Teacher Model

The process begins with a large, pre-trained model (the teacher). This model has already learned to perform tasks like language translation, image recognition, or text generation with high accuracy.

Step 2: Transferring Knowledge

Next, the teacher model’s knowledge is transferred to a smaller student model. This is done by training the student to replicate the teacher’s outputs, often using a technique called soft labels. Unlike hard labels (e.g., “this is a cat”), soft labels provide probabilities (e.g., “80% cat, 20% dog”), giving the student more nuanced information to learn from.

Step 3: Fine-Tuning the Student Model

Once the student model has learned the basics, it’s fine-tuned on specific tasks to improve its performance. The result is a compact, efficient model that performs nearly as well as the original teacher model but with significantly fewer resources.

Why AI Distillation Matters

Benefits of AI Distillation

Efficiency: Smaller models require less computational power, making them ideal for real-time applications.
Cost-Effectiveness: Reduced resource requirements lower operational costs.
Accessibility: Distilled models can run on devices with limited processing power, expanding AI’s reach.
Sustainability: Lower energy consumption makes AI more environmentally friendly.

Real-World Applications

Healthcare: Distilled AI models can analyze medical images on portable devices, enabling faster diagnoses in remote areas.
Autonomous Vehicles: Smaller models allow self-driving cars to process data in real time without relying on cloud servers.
Customer Service: Chatbots powered by distilled models provide faster, more accurate responses.

DeepSeek’s Rise: How AI Distillation Gave Them an Edge

The David vs. Goliath Story

In the AI industry, OpenAI has long been the dominant player, with its GPT models setting the standard for natural language processing. But in 2023, a relatively unknown company called DeepSeek made headlines by releasing a distilled AI model that outperformed OpenAI’s offerings in several benchmarks.

DeepSeek’s Secret Sauce

DeepSeek’s success can be attributed to its innovative use of AI Distillation. Here’s how they did it:

1. Focusing on Specific Use Cases

While OpenAI’s models are designed to be general-purpose, DeepSeek focused on optimizing for specific tasks, such as legal document analysis and financial forecasting. By tailoring their models to niche applications, they achieved higher accuracy with fewer parameters.

2. Leveraging Advanced Distillation Techniques

DeepSeek employed state-of-the-art distillation methods, including multi-teacher distillation (using multiple teacher models to train a single student) and self-distillation (where the model teaches itself). These techniques allowed them to create highly efficient models without sacrificing performance.

3. Prioritizing Accessibility

DeepSeek made their models open-source, allowing developers worldwide to experiment and build upon their work. This move not only fostered innovation but also positioned DeepSeek as a leader in the AI community.

The Impact of DeepSeek’s Breakthrough

Shaking Up the AI Industry

DeepSeek’s success has forced established players like OpenAI to rethink their strategies. By proving that smaller, more efficient models can compete with giants, DeepSeek has sparked a wave of innovation in AI optimization.

Empowering Smaller Players

DeepSeek’s open-source approach has leveled the playing field, enabling startups and researchers to access cutting-edge AI technology without the need for massive budgets.

Setting New Standards

DeepSeek’s models have set new benchmarks for efficiency and performance, pushing the entire industry to prioritize sustainability and accessibility.

Expert Insights: What the Future Holds

We reached out to Dr. Emily Carter, a leading AI researcher, to get her take on AI Distillation and DeepSeek’s impact.

“AI Distillation is a game-changer. It allows us to democratize AI, making it accessible to a wider range of applications and users. DeepSeek’s approach is particularly impressive because they’ve shown that you don’t need billions of parameters to achieve state-of-the-art performance. This is the future of AI.”

How to Leverage AI Distillation in Your Business

Step 1: Identify Your Needs

Determine which tasks could benefit from AI optimization. For example, if you’re in e-commerce, a distilled model could improve product recommendations.

Step 2: Choose the Right Tools

Explore open-source frameworks like TensorFlow or PyTorch, which support AI Distillation techniques.

Step 3: Collaborate with Experts

Partner with AI specialists to design and implement distilled models tailored to your business.

The Future of AI is Small

AI Distillation is more than just a technical innovation—it’s a paradigm shift. By making AI faster, cheaper, and more accessible, it has the potential to transform industries and improve lives.

DeepSeek’s success story is a testament to the power of innovation and strategic thinking. By leveraging AI Distillation, they’ve not only challenged OpenAI but also paved the way for a more inclusive and sustainable AI ecosystem.

As we look to the future, one thing is clear: the era of bloated, resource-hungry AI models is coming to an end. The future belongs to lean, efficient systems that deliver maximum impact with minimal resources. And thanks to pioneers like DeepSeek, that future is closer than ever.

News

Zoox’s Robotaxi: A Glimpse into the Future of Autonomous Mobility

by George Mensah 2025-02-16

Imagine stepping into a vehicle that looks like it belongs in a sci-fi movie—no steering wheel, no traditional front or back, and doors that slide open to reveal a futuristic interior. This isn’t a scene from a Hollywood blockbuster; it’s the reality of Zoox, Amazon’s autonomous vehicle subsidiary, which is quietly revolutionizing the future of transportation.

During this year’s CES, I had the opportunity to experience Zoox’s second-generation robotaxi firsthand. This purpose-built, bidirectional electric vehicle is designed to operate without human intervention, and it’s already being tested on the streets of Las Vegas. But is it ready for prime time? Let’s dive into the details.

What is Zoox?

Zoox is an autonomous vehicle company acquired by Amazon in 2020. Unlike other self-driving car projects that retrofit existing vehicles, Zoox is building its robotaxis from the ground up. The result is a unique, box-shaped vehicle that’s unlike anything on the road today.

Key Features of the Zoox Robotaxi:

Bidirectional Design: The vehicle can move forward or backward without turning around, making it highly maneuverable in urban environments.
No Steering Wheel: Fully autonomous, with no traditional controls.
Touchpad Controls: Passengers interact with the vehicle via tablets mounted inside.
Spacious Interior: Seats face each other, creating a social, subway-like atmosphere.

The Road to Autonomy: Zoox’s Journey

Zoox has been working on its autonomous technology for over a decade. While competitors like Waymo (owned by Alphabet) have already launched public robotaxi services, Zoox has taken a more cautious approach.

Current Status:

Testing Locations: Zoox is testing its vehicles in Las Vegas, San Francisco, Seattle, and soon Austin and Miami.
Limited Access: For now, the service is only available to employees and their families.
Public Launch: Zoox plans to open its robotaxi service to the public in Las Vegas by 2025, though details about waitlists and availability remain unclear.

Inside the Zoox Robotaxi: A Firsthand Experience

Stepping into the Zoox robotaxi feels like entering a futuristic pod. The interior is minimalist yet functional, with firm, easy-to-clean seats and custom airbags for safety. Here’s what stood out during my ride:

1. Comfort and Space

The vehicle’s low floor and sliding doors make it easy to enter and exit. The seats face each other, creating a communal atmosphere, though sitting backward might not be ideal for those prone to motion sickness.

2. Controls and Features

Passengers interact with the vehicle via tablets mounted near the doors. These controls allow you to adjust the temperature, request a stop, or check the ride duration. While functional, the interface feels basic compared to the sleek designs of modern car infotainment systems.

3. The Ride Experience

The 30-minute test loop took us through local streets near the Las Vegas Strip. The vehicle handled construction zones, pedestrians, and traffic lights with ease, though its decision-making felt cautious and slightly robotic.

Highlights:

Construction Zone Navigation: The robotaxi successfully navigated a lane closure, though it took longer than a human driver might.
Pedestrian Interaction: It waited patiently for a pedestrian to cross, demonstrating its ability to handle real-world scenarios.

Challenges:

Jerky Braking: The vehicle occasionally braked abruptly, which could be unsettling for passengers.
Rough Ride: The low-slung design and minimal suspension made potholes and rough roads feel more pronounced.

How Zoox Compares to Competitors

Zoox is entering a crowded field, with players like Waymo, Cruise, and Tesla vying for dominance in the autonomous vehicle space. Here’s how it stacks up:

1. Waymo

Advantage: Waymo has a head start, with public robotaxi services already operating in Phoenix and San Francisco.
Challenge: Waymo uses retrofitted vehicles, which lack the custom design of Zoox’s purpose-built robotaxis.

2. Cruise

Advantage: Cruise has made significant strides in autonomous technology but faced setbacks, including accidents and regulatory scrutiny.
Challenge: Cruise recently paused its operations to address safety concerns, highlighting the difficulties of scaling autonomous services.

3. Tesla

Advantage: Tesla’s Autopilot and Full Self-Driving (FSD) features are widely used, though they still require human supervision.
Challenge: Tesla’s approach to autonomy has been criticized for prioritizing speed over safety.

The Challenges Ahead for Zoox

While Zoox’s technology is impressive, several hurdles remain before it can achieve widespread adoption:

1. Regulatory Approval

Autonomous vehicles must meet stringent safety standards, and Zoox’s self-certification process has raised questions among regulators.

2. Public Trust

Convincing passengers to trust a fully autonomous vehicle will require flawless performance and transparent communication.

3. Scalability

Building and deploying a fleet of custom vehicles is expensive and labor-intensive. Zoox will need to prove that its business model is sustainable in the long term.

Why Zoox Matters

Despite the challenges, Zoox represents a bold vision for the future of transportation. Its purpose-built design and focus on safety set it apart from competitors, and its backing by Amazon provides the resources needed to scale its operations.

As Zoox cofounder and CTO Jesse Levinson put it: “This experience that we’re having in this vehicle, there’s nothing like it now. You can’t get that in a retrofitted car. We think we’re onto something.”

The Future of Mobility is Here

Zoox’s robotaxi is more than just a vehicle—it’s a glimpse into a future where transportation is safer, more efficient, and more accessible. While the road to full autonomy is fraught with challenges, Zoox’s innovative approach and steady progress suggest that the future is closer than we think.

So, the next time you’re stuck in traffic, imagine a world where your commute is handled by a sleek, autonomous pod. Thanks to Zoox, that future might be just around the corner.

Load More Posts