A team of researchers at Georgia Tech has developed a new smartphone-based system that could dramatically simplify how people interact with robots. Called COBALT, the platform allows users with little to no computing experience to remotely control robot arms from virtually anywhere in the world using just a phone and an internet connection.

The project, developed at Georgia Tech’s People, AI & Robotics (PAIR) Lab, transforms smartphones into motion controllers for robotic arms. Users simply move their phones in different directions, and the robot mirrors those movements in real time. Basic tasks such as grabbing, moving, and releasing objects can be performed through simple on-screen controls, making the experience feel more like playing a mobile game than operating industrial machinery.

Ayush Agarwal, a Ph.D. student in Georgia Tech’s School of Interactive Computing who leads the COBALT research team, said the system was intentionally designed to make robotics accessible to beginners rather than experts. During testing, participants from countries including India, Indonesia, and Pakistan remotely controlled robot arms located inside Georgia Tech’s lab despite having no prior robotics experience.

Researchers believe crowdsourcing could shape the future of robotics

The broader goal behind COBALT extends beyond convenience. Researchers believe the platform could solve one of robotics’ biggest challenges: collecting enough real-world training data to improve AI-powered robotic systems.

Modern robots require enormous amounts of policy training data to learn how to perform physical tasks reliably. According to Assistant Professor Animesh Garg, who directs the PAIR Lab, simulation alone is not enough to train robots for large-scale deployment. Instead, researchers envision a crowdsourced network where millions of smartphone users passively contribute operational data by remotely interacting with robots.

Garg compared the idea to tapping into the nearly five billion smartphone users worldwide. By lowering the barrier to entry, the team hopes to create a scalable global system capable of accelerating robotic learning and automation.

The technology could also have major educational implications. Georgia Tech researchers recently demonstrated COBALT to students from Midtown High School in Atlanta, allowing them to remotely operate robot arms using smartphones. The simplicity of the interface could make robotics education more accessible in classrooms without expensive equipment or specialized hardware.

A future “gig economy” for robots may not be far away

The researchers also believe COBALT could eventually support entirely new forms of remote work. Garg described the possibility of a robot-powered gig economy where people remotely operate assistive robots in homes, warehouses, or factories from anywhere in the world.

In practical terms, that could mean a factory robot autonomously handles most tasks but requests human assistance when it encounters a difficult situation. Instead of requiring on-site workers, remote operators could briefly take control through their phones before handing the operation back to the AI system.

Agarwal said user studies showed smartphones were preferred over VR headsets, keyboards, or traditional controllers because they felt more intuitive while still providing high-quality control data. The system also minimizes latency by using WebRTC technology, similar to platforms like Zoom and Google Meet, ensuring that robot movements and live video streams remain responsive even across long distances.

The research paper on COBALT is being presented this week at the IEEE International Conference on Robotics and Automation in Vienna, where the team is showcasing not just the technology itself, but the large-scale remote operation network built around it.

Tesla has officially pushed back the public debut of its next-generation Roadster to August or later this year. The engineers are still ironing out the bugs in the cold gas thruster kit that they are creating with SpaceX. The event is expected to take place in Texas, and it will be Tesla’s first full-fledged vehicle showcase since the Cybercab presentation in 2024.

Apparently, an early version of the thruster system arrived at Elon Musk’s desk for review in late April. Cold gas thrusters are intended to give the Roadster greater power, faster acceleration, tighter handling, and better braking, all due to how the gas flows out of those specific nozzles. According to the desired performance criteria, once all cylinders are running, a zero-to-sixty run should be doable in roughly a second. The limited edition SpaceX vehicle will apparently include a plethora of thrusters spread throughout, ten in total, plus the fuel to power them, all squeezed into the space normally occupied by the back seats. A standard version with fewer thrusters and tanks will likely be available to the general public.

The first customer automobiles are expected to leave the Texas Gigafactory in either 2027 or 2028, according to the most recent production estimates. Anyone who put down a deposit back in 2017 is still waiting for their automobile, which must be frustrating.

They appear to have made multiple timeline modifications throughout the years; twelve years after the first idea in November 2017, they are still working on it. Customer deliveries were originally scheduled for 2020, however the timeframe was delayed multiple times. It was previously expected that the demo date would be April 1, followed by late Spring, but it has now been pushed out to late Summer. Elon stated on the April earnings call that the event could be just a month or so away, but it now appears to be a pipe dream.

The main cause of the delays is ongoing development on the A71 thruster system, which is a cold gas thruster configuration in which the gas is simply pushed via nozzles rather than burned up, as in a conventional rocket engine. This makes the hardware easier and safer to drive on the road while yet offering adequate performance. In recent weeks, several trademark applications have been filed for the new Roadster design.

Franz von Holzhausen, Chief Designer, and Lars Moravy, Vice President of Engineering, have kept the program on track despite the fact that it looks to have been completed many years ago. Perhaps August will explain how all of these components, including Tesla’s unique electric engine, thrusters, and aerodynamics package, will work together.
[Source]

from the ctrl-alt-speech dept

Ctrl-Alt-Speech is a weekly podcast about the latest news in online speech, from Mike Masnick and Everything in Moderation‘s Ben Whitelaw.

Subscribe now on Apple Podcasts, Overcast, Spotify, Pocket Casts, YouTube, or your podcast app of choice — or go straight to the RSS feed. To get extended episodes with additional coverage, support us on Patreon.

In In this week’s episode, Mike and Ben cover:

And in the extended episode for Patreon supporters, they cover:

Ctrl-Alt-Speech is the podcast where we make sense of the major debates shaping online speech, platform power, content moderation and the future of the internet. It’s co-hosted by Mike Masnick (Techdirt) and Ben Whitelaw (Everything in Moderation).

If you’re already a Patreon supporter, you can get the extended episode on Patreon.

Filed Under: ai, artificial intelligence, content moderation, decentralization, trust and safety

Companies: bricks and minifigs, meta

from the none-of-this-makes-sense dept

Full disclosure: this post is going to pose way more questions than answers. That’s because the story of the Tomb Raider remake being produced by Crystal Dynamics and its inclusion of an AI disclosure on Steam makes no sense to me.

So, let’s start at the beginning. Crystal Dynamics is making an updated version of the first Tomb Raider game and it looks pretty great from what I’ve seen. But, as gamers are now accustomed to doing, the public came to notice that the game’s Steam page included one of Steam’s mandatory AI disclosure notices. It reads thusly:

AI-assisted tools were used during development to support some early exploration and temporary development content. Any AI-assisted assets were either replaced or refined by humans in order to maintain the creative and artistic vision of the development team.

And from there, because, of course, everyone freaked out. Comments from all the corners of the internet began flooding in, swearing off ever touching this game because it was developed using AI. What AI? We don’t know. How much was it utilized? No real content there, either. But is the game going to be good? It doesn’t fucking matter, because AI was used and that’s all you need to know in order to know that this game is going to be artless pap fit only to be mocked and laughed at.

Even some gaming journalists have gotten into the habit. This is from Kotaku:

GenAI slop potentially showing up in Tomb Raider is disappointing but maybe less surprising than it should be. Phil Rogers, CEO of Crystal Dynamics’ parent company, Embracer Group, last year called genAI a “powerful technology” for “driving efficiency.” Crystal Dynamics has also undergone several rounds of layoffs, completing three just last year and one earlier in 2026.

Here you have a writer who has already reached their conclusion while having almost zero information on which to base that conclusion. They haven’t played the game. They’ve barely seen the content in the game, save for some trailers. They don’t know thing one about how AI was used, where, and in what way. But it’s probably going to be “GenAI slop”. As if there is simply no other possible outcome.

But it gets at least slightly stranger with Crystal Dynamics having responded to some of those concerns over at Eurogamer.

“At Crystal Dynamics, we leverage AI tools to help our teams iterate on ideas faster and more efficiently, while ensuring that all finished content in the final product is human-crafted. Our goal is to empower the creativity and flexibility of our developers to deliver the highest-quality experiences for players everywhere.”

This is where I get confused. If all of the content that is going to make it into the final product is “human-crafted,” then they shouldn’t even have needed to add the disclosure to their Steam page. Back in January, Steam updated its rules around its AI disclosure such that a game with an AI disclosure must have AI-generated content that is either in public marketing materials for the game or in the final product and with which the player of the game interacts in order to require the disclosure.

In its submission form, Valve now specifies that game publishers must disclose pre-made generative AI assets only when used in marketing materials or content that “ships with your game, and is consumed by players.”

In other words, Steam’s disclosure requirement is not concerned with generative AI tools used behind the scenes for efficiency gains (presumably including coding helpers) or office work, but with things like final art, sound, and writing.

Advertisement

Now, I’ll just note that there is a subtle difference in the disclosure notice and Crystal Dynamics’ statement. The former indicates that the game may include AI-generated assets that were then iterated upon by a human developer. The latter seems to say the opposite, where everything in the final game will be “human-crafted”. So… which is it?

As I said from the start, more questions than answers is all I have at the moment. But if the gaming public is going to freak out at the mere mention of some AI being used in some way, somewhere within every new video game that comes out, then this is going to be a very annoying time in which to be a gamer.

Filed Under: ai, tomb raider, video games

Companies: crystal dynamics

A Chinese espionage group tracked as UNC5221 has been accessing Microsoft 365 environments using the Brickstorm backdoor and previously undocumented malware named Plenet and AgentPSD.

An investigation into the incident revealed that the threat actor had gained access to the victim network at least 18 months before detection, and had also compromised the victim organization’s managed services provider (MSP).

UNC5221 is also tracked as VerdantBamboo and has been involved in attacks that exploited zero-day vulnerabilities in edge devices since at least 2023.

The threat actor used the Brickstorm backdoor undetected in the environments of various targets in the United States for more than a year until the breaches were discovered around March 2025.

Researchers describe Brickstorm as “an advanced malware implant.” Initial variants were written in Golang, then new variants emerged, written in Rust.

In April 2024, Google documented UNC5221 activity using the backdoor, and then again in September 2025, describing attacks against legal services, software-as-a-service providers, business process outsourcers, and technology companies.

CISA warned about Brickstorm being deployed by Chinese hackers against VMware vSphere servers, and, more recently, Google reported that it was deployed by UNC6201 against Dell RecoverPoint for Virtual Machines.

Victim hacked twice

Volexity researchers responding to an incident last year found that VerdantBamboo compromised an Egnyte Storage Sync system and accessed it periodically through the victim’s web SSL VPN.

From this foothold and using Brickstorm proxying features and stolen credentials, the threat actor accessed the organization’s Microsoft 365 enevironment.

“Volexity assesses with high confidence that this was done to blend in with legitimate network traffic and evade Conditional Access policies that would have otherwise prevented access,” the researchers said.

Later, Volexity discovered that the hackers had spent at least 18 months on the network before being detected. Furthermore, VerdantBamboo breached the organization again after the researchers completed the remediation efforts.

In the second intrusion, the attackers used stolen credentials to enable and configure SSL VPN access on the victim’s firewall, then connected to internal systems and deployed additional custom malware to a Synology NAS device.

This triggered an investigation at the customer’s MSP, where Volexity found that VerdantBamboo had planted a BSD variant of Brickstorm on a pfSense firewall.

“Volexity concluded that this firewall, like the victim organization’s Storage Sync system, had also been compromised at least 18 months earlier.”

The researchers have medium confidence that the attacker pivoted from the MSP into the victim organization’s environment.

Brickstorm was then deployed to the victim’s Egnyte Storage Sync appliance and to a retired Linux GroupWise email archive server.

New backdoors used

Once the attackers returned a few days later and re-established access to the victim’s infrastructure, they deployed the custom malware Plenet to a Synology NAS appliance.

Plenet, also tracked as “Grimbolt” by Google, is a cross-platform .NET-based backdoor that offers interactive shell access, remote command execution, file manipulation, and command-and-control (C2) server switching.

The researchers note that Plenet is similar in design to Brockstorm, using the WebSocket protocol for C2 communications and a multiplexing library for simultaneous data streams to the server.

AgentPSD is a simple Python-based reverse shell utility that Volexity believes VerdantBamboo used as a fallback persistence mechanism if other malware was no longer accessible.

The researchers discovered that AgentPSD was configured to connect to a different domain than the one Brickstorm used. However, the malware was never used as Brickstorm was still running, which supports the assessment that AgentPSD was a secondary access mechanism.

During the investigation, Volexity tried to discover the infrastructure related to VerdantBamboo. The researchers created a fingerprint to identify IP addresses and domains Brickstorm used for C2 communication.

Although multiple machines were identified, the threat actor took the infrastructure offline before the researchers could reveal other systems.

“Between September 18 and September 23, all of the servers previously matching this pattern turned off their services on port 443.”

Around that time, Google also published a new report on Brickstorm’s activity, which may suggest that the attacker was aware of their operations being under investigation.

Volexity’s describes VerdantBamboo/UNC5221 as “a highly sophisticated threat actor” that mixes living-off-the-land techniques and malware and targets systems that do not support endpoint detection and response (EDR) solutions.

The researchers compiled a list of indicators of compromise (IOCs) linked to the investigated UNC5221 campaign and published them here.

Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.

The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.

Get the whitepaper