Blog
Dr. Ryan Ries here, back with this week’s Mission Matrix.
Today I have 3 news stories that I found interesting, and I haven’t seen a lot of people talking about them.
Proceed with caution, two of these stories are a bit… creepy.
China's Robot Marathon
This week, China organized what might be the most ambitious and scary public test of humanoid robotics we've seen yet: a half-marathon featuring 21 robot models competing alongside thousands of human runners.
The star of the show was Tien Kung Ultra, a 5'9" humanoid developed through a collaboration between UBTech, Xiaomi, and the Beijing government. This robot finished the 13-mile course in just under 2 hours and 41 minutes - impressive, but still well behind the human winner's time of about 1 hour.
I don’t know if I’m more intrigued or creeped out by this. Either way, the good news is that some of us are still able to outrun these robots!
Progress & Limitations
This event exposed both the progress and limitations of current humanoid technology.
While some robots could maintain balance and navigate the course, others barely made it past the starting line.
One robot named Shennong crashed straight into a fence, breaking into pieces. Clearly, he did not train hard enough for this half-marathon!
Practical challenges also came into play. Battery limitations forced robots like Tien Kung Ultra to make multiple swaps during the race, metal components had to replace plastic parts for durability, and most robots needed human handlers nearby.
While we're making incredible strides, we're still far from the humanoid helper robots that science fiction has promised us.
ChatGPT's "Reverse Location Search" Raises Serious Privacy Concerns
OpenAI just released their newest models — o3 and o4-mini — with enhanced visual reasoning capabilities, and users have quickly discovered something concerning: these models are remarkably good at identifying locations from photos.
People are uploading images to ChatGPT and asking it to play "GeoGuesser," essentially deducing where a photo was taken based on visual clues.
The results can be unnervingly accurate, with the AI successfully identifying specific restaurants, neighborhoods, and landmarks from subtle details.
While some might see this as a cool party trick, there are some serious privacy implications here.
There's nothing preventing someone from taking a screenshot of your Instagram story and using ChatGPT to determine your location, essentially enabling a form of AI-powered doxxing.
What's particularly interesting is that OpenAI doesn't address this issue in their safety report, suggesting they might not have anticipated this use case. After TechCrunch published an article on the trend, OpenAI issued a statement claiming they've implemented safeguards against identifying private individuals and sensitive information.
This highlights something I've been emphasizing in our client engagements: emerging AI capabilities often create unexpected privacy and security vulnerabilities that organizations need to prepare for proactively rather than reactively.
LLMs Meet Biology: A Breakthrough in Single-Cell Analysis
Okay, let's move on to a more optimistic side of AI.
Researchers have created an innovative framework called C2S-Scale that uses Large Language Models for analyzing single-cell RNA sequencing data.
The approach transforms gene expression data into "cell sentences" — sequences of gene names ordered by expression level.
This allows standard LLM architectures to process complex biological data without requiring custom modifications.
By scaling their models up to 27 billion parameters and training on over one billion tokens of transcriptomic data and biological text, the research team achieved remarkable improvements in both predictive and generative capabilities.
Their framework outperforms both specialized single-cell models and general-purpose LLMs across diverse tasks.
What's most impressive is how this approach can generate biological insights in plain language, making complex cellular biology more accessible to researchers.
It's a perfect example of what I like to tell our customers: the most transformative AI applications often come from adapting existing frameworks to new domains rather than building everything from scratch.
What's Next
Let me know your thoughts on these news stories and where AI is headed. It’s honestly amazing how far we’ve seen AI come in just the first few months of 2025.
Next week, Jonathan LaCour (Mission CTO) and Caitlin Berger (Senior Solutions Architect) will be hosting our next Generative AI “Ask Us Anything” on April 29th. If you have questions around generative AI, this is a great opportunity to ask experts in the field.
Until next time,
Ryan
Now, for this week’s AI-generated image and the prompt I used to create it. You have probably seen this trend going around if you’re on LinkedIn.
"Create a high-quality 3D illustration of an action figure toy called "Dr. Ryan Ries." The figure is displayed inside a clear, cartoon-style blister package, just like you'd find on a toy store shelf. The packaging is minimalist, with natural cardboard colors and a cute, collectible vibe. * The action figure should closely resemble the person in the reference photo, capturing his features: Brown hair, FACE SHAPE, EYE COLOR, SKIN TONE, ETC.]. * At the top of the packaging, in bold white letters, write: "Dr. Ryan Ries." Directly underneath, add the title "Chief AI & Data Scientist." * Dress the figure in a black tshirt with the AWS logo on the chest and denim jeans and black tennis shoes, reflecting their unique style or profession. * Include fun, job-related accessories or props next to the figure, such as anything related to AI and legos. * The overall style should be cartoonish and cute, but with clean, professional lines—think modern collectible toys. * Place the Mission, a CDW company logo in the top right corner of the packaging. I have attached this logo. Make sure the final image is vibrant, playful, and ready for the toy aisle!"
Sign up for Ryan's weekly newsletter to get early access to his content every Wednesday.
Go Back
Share this post