Introduction to Construction AI
The use of Artificial Intelligence (AI) in construction is becoming increasingly popular, with companies like Safety AI and Safeguard AI developing technologies to improve site safety. These companies use vision language models (VLMs) and machine learning to analyze images and videos from construction sites, identifying potential hazards and alerting site managers.
How Construction AI Works
Safety AI’s system uses a combination of VLMs and older machine-learning methods to analyze images and videos from construction sites. The system has over a dozen layers of questioning to determine whether a ladder is being used safely, for example. The company’s CEO, Lorenzo, hopes to have his methodology independently audited by safety experts to ensure its accuracy.
The Importance of Accuracy
While Safety AI’s system has shown promise, with a success rate of 95%, there are still concerns about its accuracy. Chen Feng, who leads New York University’s AI4CE lab, notes that VLMs have "some pretty fundamental issues" to resolve, including hallucinations and edge cases. These are anomalous hazards that the VLM hasn’t trained on, and they can be difficult to detect.
The Challenges of Edge Cases
Feng points to a 2024 paper called "Eyes Wide Shut?" that noted "systematic shortcomings" in VLMs. The paper highlights the challenges of interpreting 3D scene structure from 2D images, as well as the lack of situational awareness and common sense in VLMs. These challenges can make it difficult for VLMs to detect edge cases, which can be a major flaw in the system.
Combining Old and New Methods
To address these challenges, Safety AI combines VLMs with older machine-learning methods, such as segmentation and photogrammetry. The company has also trained its system on 10 different problem areas, including ladder usage, to anticipate common violations. While the system is not perfect, it can still provide an extra set of "eyes" for overworked safety managers.
The Benefits of Construction AI
Aaron Tan, a concrete project manager, notes that a tool like Safety AI could be helpful for safety managers, who are often responsible for multiple sites at once. The system can save time and improve safety by providing emailed alerts rather than requiring site visits. However, Tan also notes that workers may be resistant to the technology, fearing it will be used to monitor their activities.
Older Methods Still Have Value
Izhak Paz, CEO of Safeguard AI, has chosen to stick with older machine-learning methods, which he considers more reliable. The company’s system uses a hybrid approach, combining machine learning with human intervention to deal with deviations. The system can be trained on new categories of danger, and it can provide real-time risk assessments and instructions to site managers.
The Cost of Construction AI
While Safeguard AI’s system is not suitable for small-scale builders, it is affordable for mid-market and larger builders who manage multiple sites. The system is currently in use at roughly 3,500 sites in Israel, the United States, and Brazil.
Conclusion
Construction AI has the potential to improve site safety and reduce the risk of accidents. While there are still challenges to be addressed, companies like Safety AI and Safeguard AI are making progress in developing accurate and reliable systems. By combining old and new methods, these companies can provide valuable tools for safety managers and site workers.
FAQs
- What is Construction AI?
Construction AI refers to the use of Artificial Intelligence (AI) and machine learning to analyze images and videos from construction sites, identifying potential hazards and alerting site managers. - How does Safety AI’s system work?
Safety AI’s system uses a combination of VLMs and older machine-learning methods to analyze images and videos from construction sites, identifying potential hazards and alerting site managers. - What are the challenges of using VLMs in construction AI?
VLMs have "some pretty fundamental issues" to resolve, including hallucinations and edge cases, which can be difficult to detect. - How can Construction AI benefit site workers and safety managers?
Construction AI can save time and improve safety by providing emailed alerts rather than requiring site visits, and it can provide an extra set of "eyes" for overworked safety managers. - What is the cost of Construction AI?
The cost of Construction AI varies, but Safeguard AI’s system is affordable for mid-market and larger builders who manage multiple sites.
