This video has three parts:
1) Professor Kornhauser’s comments on a podcast that were turned into a video about an autonomous delivery vehicle with a unique way of delivering packages.
2) How AI may eventually be a new form of compression for images and video.
3) How this may eventually lead to the science fiction Vulcan mind meld.1
Read more below.
The often-overlooked value of artificial intelligence lies in its ability to compress information. Entire images or complex, time-based video clips can be described using just a handful of words.
This concept was central to a February 2025 Viodi article, which documented early experimentation with AI-driven data compression. The premise is simple: transmitting a short textual description requires vastly lower bitrates than sending raw video files using traditional formats like JPEG or MP4. While traditional compression sacrifices visual sharpness to stay faithful to the original pixels, generative AI flips the model: it yields sharp, convincing visuals by interpreting intent, even if the exact fine details diverge from the original source.
That brings us to a recent episode of Professor Alain Kornhauser’s SmartDrivingCar podcast. During the discussion, Alain jokingly proposed a unique solution to the “last 50 feet” problem in autonomous goods delivery: an autonomous delivery vehicle equipped with a turret that shoots packages directly onto front porches. His verbal description, carried over modern telecom networks, instantly created a clear mental image in my mind.
Later that evening, I fed the following prompt into Grok:
“An autonomous delivery vehicle that solves the last 50 feet of package delivery by shooting the package from a turret that is on top of the autonomous delivery vehicle. The packages fly through the air and land on the front porches.”
On the very first try, Grok generated a complete video.
While the output had minor imperfections and didn’t match our exact internal mental pictures, it captured probably 90% of what we both envisioned. What is truly remarkable is the efficiency: a text prompt of just 254 bytes generated a 7.6-megabyte video file.
The Network Implications #
The broader implication for telecommunication networks is profound: required bandwidth for human-consumed content could drop significantly in the future. Instead of streaming heavy video files over the core network, AI at the edge—embedded in homes, field cabinets, or central offices—will reconstruct video, audio, and other sensory media locally using tiny text seeds or semantic instructions. (Though agent-to-agent communication may present entirely new challenges to the broadband network).
As brain-computer interfaces, wearables, and implants advance toward direct read-write access, these signal and interpretation losses will diminish. When that happens, the informal “Vulcan mind meld” Alain and I experienced will be perfected—paving the way toward true superintelligence.
Note: the author wrote this article with editing assistance from Gemini.
- The autonomous vehicle was generated by Grok (first try). The compression and Vulcan mind meld scenes were produced by Gemini. ↩︎
Leave a Reply