MIT’s DAAAM gives robots long-term memory by attaching language descriptions to 3D maps. You can ask “where did I leave my wallet?” and it knows.
Robots are still surprisingly bad at remembering where things are. You might recall that your keys were on the kitchen counter last night. A robot working beside you would struggle to connect that object and location in a useful way. MIT researchers built a system called DAAAM to fix that.
DAAAM stands for Describe Anything, Anywhere, Anytime, at Any Moment. It combines computer vision and 3D mapping to give robots a long-term spatial memory. As a robot moves through an environment, it attaches detailed language descriptions to objects it sees and stores them in a spatial map. Instead of just knowing there is an object at a coordinate, it remembers that there is a red bicycle with a flat tire near a specific building.
A person can then ask natural language questions: “Where did I leave my wallet?” or “Go grab the component we started assembling last night.” The robot searches its memory for the right object and location. The system runs fast enough for a mobile robot to use in real time.
The researchers found DAAAM answered questions more accurately than current methods, depending on the query type. The work was presented at the Conference on Computer Vision and Pattern Recognition (CVPR) and is available as a preprint on arXiv.
The system is not ready for consumer products. It is a research framework that shows what is possible when you combine vision, language, and 3D spatial data into a persistent memory layer. The researchers are still working on giving the system better confidence levels and helping it remember significant events, not just static object placements.
The gap DAAAM addresses is fundamental to useful robotics. Physical AI systems need to understand the real world, not just process text. A robot that can clean a house, manage a warehouse, or assist in a factory needs to know not just what it sees right now, but what it saw yesterday and where. Current robots either forget everything between tasks or require expensive pre-mapping of every environment.
DAAAM’s approach is practical because it does not require the environment to be set up in advance. The robot builds its memory as it moves. MIT has been publishing a series of robotics breakthroughs this year, including an ultrasound wristband for remote robot control. DAAAM tackles the other side of the problem: not how to control a robot, but how to make it remember what it has seen. Intelligence without memory is not intelligence. It is reaction.
Staff who use AI can end up with more to do, not less.
Think carefully about the tools you’re using and why.
Adopt a set of standards and refine your outputs.
The promise of productivity boosts from AI can come with an unwelcome side order of stress. Harvard Business Review found that AI doesn’t reduce work; it intensifies it, leading to cognitive fatigue and unsustainable hours.
While the common perception is that AI can help reduce workloads, allowing employees to focus more on higher-value and more engaging tasks, HBR’s research found that staff using AI worked more quickly and often ended up with more to do, not less.
Ankur Anand, group CIO at tech recruiter Harvey Nash, said professionals who want to avoid cognitive fatigue must understand how to use AI effectively and its potential risks.
“That focus will help to reduce the noise around the workload that AI creates,” he told ZDNET, suggesting that many people have unrealistic expectations about the productivity boost that AI will provide.
“Many organizations are telling their people, ‘We want to understand how you’re making an impact with AI,'” he said. “But these professionals are not empowered, which means that using AI adds a lot of pressure, because they need to prove themselves on their own terms.”
If you’re going to make the most of AI at work, then you’re going to have to find an effective balance between completing tasks quickly and producing high-quality work.
Here’s how the experts believe professionals can ensure they reap the benefits, not the problems, of AI — and they suggest that you’ll need to focus on three core areas: tools, guidelines, and outputs.
Limit your toolset
Alex Read, senior enterprise product manager for data at energy provider EDF UK, told ZDNET that the best way for professionals to reap the benefits, not the challenges, of AI is to be uber-focused on tools that help you produce value in your roles.
While there are thousands of potential AI-enabled services on the market, Read said sensible professionals limit their horizons.
In his own role, for example, Read focuses on how AI can help him build a data platform and update information accurately, efficiently, and productively: “Anything outside of that scope is noise for me.”
That sentiment resonated with Nick Pearson, CIO at technology specialist Ricoh Europe, who told ZDNET it’s important to take a step back and think carefully about how an AI tool can help you produce value in your role.
“If you think about the phrase ‘gen AI,’ the tech is very good, by definition, at generating outputs,” he said. “I could go to bed in the evening, set the model to work, and we could have four new IT strategies produced overnight.”
However, quantity doesn’t necessarily mean quality. Pearson suggested it’s important to focus on AI’s blind spots, particularly as most models are trained on preexisting content.
“AI can’t inspire people, per se; it can’t naturally create something new, because it’s actually quite recursive,” he said.
“And the judgment you have to put in sometimes, on top of everything else, whether it be an ethical or a capability judgment, is not there automatically in the technology.”
It’s in this gap, said Pearson, that human experts play a critical role: “We’re toying with that concern as an organization and saying, ‘Where does AI really play an important role, versus where are we upskilling people in areas that AI probably won’t play for a long time?'”
To correct this issue, HBR said companies need to adopt an “AI practice,” or a set of norms and standards around AI use that help professionals ensure they use AI in a constrained but productive manner.
At EDF UK, Read is part of an internal AI Center of Excellence in enterprise IT, which enables policy for the effective use of AI across the wider organization.
In addition to Read, who contributes input from a data-use perspective, the group includes other tech representatives, such as the firm’s senior manager of AI, principal software engineer, and principal solution architect.
“The remit of this center is to make sure that, when the federated business units are looking to build, develop, and deploy AI services, they have platforms, guidance, best practices, architectural assets, and materials to guide them on how to safely and efficiently adopt AI and operationalize it at scale,” he said.
Some of the key themes the center considers when assessing AI tools are scalability and reusability, ensuring a proposed service doesn’t replicate one already in use.
“All new tools and services related to AI will go through that hopper and funnel to understand scope and ensure the security, regulatory, and ethical side of things are understood,” he said, suggesting that all professionals should use their organization’s pre-existing guidelines to foster an appropriate exploitation of emerging tech.
“The benefit that guided approach brings is that it allows us to be clear in our messaging around what AI services can be used, how they’re used from a use-case perspective, and ultimately, what personas are allowed to use them.”
Louise Newbury-Smith, head of UK&I at technology specialist Zoom, told ZDNET that one way to ensure your outputs are constrained is to focus on prompting.
“Use simple amendments to be specific, such as ‘Give me the top three things with the biggest impact.’ That approach should guide your prompt, rather than saying, ‘Give me everything you know about this topic.'”
Newbury-Smith said the successful use of AI is all about being smart about how it’s exploited, and that effectiveness comes down to enablement and engagement. If a prompt yields too much information, refine it until you get what you need. She said this should still be faster than trying to get answers without AI.
The basic message for professionals is that effective applications of AI are all about you staying in the loop, said Bernhard Seiser, vice president of digital, data, and IT at AOP Health.
Think before you use AI, and think again before you push your outputs around the organization.
“It doesn’t help the business if you get AI-generated emails that are many pages long, and then you need ChatGPT to summarize the text,” he told ZDNET.
Seiser said that while there are certain tasks generative AI is good at and worth using for, in the end, “you need to use your brain.”
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behavior or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.
