Last week I was privileged to be invited to an event in Sydney hosted by AMD and HP, with support from Microsoft and Blackmagic Design. The subject of the event was, of course as is the thing these days, Artificial Intelligence, or more commonly, AI.
This overview of the proceedings is strictly from my viewpoint; if you were also there, you may certainly have formed a different opinion to me and of course I welcome your input if you so wish!
My Pre-Event Thoughts
It is fair to say to date, that from a photo / video creative viewpoint, I am less than impressed with AI, as results have generally been clunky and rather than “groundbreaking”, more often than not, appear to be using algorithms to correct glaring mistakes by the creator, in their photo or video.
Alternatively, “AI” has been used to manufacture imagery that is just cobbled together bits of stuff trawled off the internet – and often with laughable results.
So, the big question is, did I come away from this with a change of mind and heart?
Before I answer that, let me briefly describe what was shown and discussed at the event.
Initial Demonstrations of AI
Before the event started proper, we were given some demos of AI solutions using HP laptops with AMD processors from Microsoft, HP and Blackmagic Design., There was also a demo using some beta software.
The Microsoft demo was using the latest version of Co-Pilot, and it was explained it was still in the testing phase so there were some issues. We were shown a text document t that was basically in invitation list to a wedding. Using Co-Pilot, the document contents were then integrated into PowerPoint – yes PowerPoint no kidding! – to create individual graphically based invites.
The imagery was pulled from Microsoft’s own stock to do this I was told, but equally you could use your own if you wished.
My thoughts? OK, its early days, but I didn’t think this was any more advanced than a mail merge template could achieve, and I certainly didn’t see any real “AI” in the process. But I may be missing something.
The next demo was also graphically based and involved the web cam in the laptop taking a photo of one of the participants and adding filters to create an Avatar based on the person’s face. A second demo let participants type a description of a scene and the software created it; for example, we saw a squadron of flying pigs, and images of wide-eyed kittens.
So, sorry, nothing really new there either.
The third was a little meatier, and demonstrated by a representative from HP. In short, a specification sheet from a laptop was analysed by AI to break down all the technical stuff into something a layman could understand. The example used was “I am a content creator, so out of all this technical information, tell me the things I need to know to show me if this suits my needs”.
Seconds later, a PDF was generated (at least it looked like a PDF but I might be wrong there) detailing what the AI thought the person making the request needed to know.
Quite impressive I admit. But, I asked the demonstrater, “how does the AI know what information I need”. Apparently, the answer to that question is all in the way you phrase how you tell the AI what you want. So, some training will be needed in that area then I would suggest, and it is not something that would come naturally to most people (as from a technical perspective they don’t know what they want in my experience, as that is not their area).
So, I am guessing AI will need a way to go as yet to make this an everyday “thing”.
Next up was a quick demonstration of the Topaz Labs software to clean up around 30 seconds of 240p video and upscale it. This was impressive, but again, I am not quite sure AI per se is used as against some, admittedly very complex, mathematical computations at a pixel and even deeper level. The AI did not actually work out or predict anything that I could see.
The final demo was one I could get my teeth into as it involved Blackmagic Design’s DaVinci Resolve and its AI integration in the Fusion module.
An image of a fantasy female in a video was shown, and using the AI, sections of her face were “corrected”. These included areas of the actual face, eyes, mouth and so on. To do this, it was explained, AI is used to analyse what is there, and then make corrections based on that knowledge to improve it.
Again, very impressive, and in this case, very fast showing the power of the AMD chip behind it all. I am mostly reserving judgement on this demo, as it is the one I can mimic back in my studio as a DaVinci Resolve user but suffice to say it does need a deep knowledge of Resolve and Fusion in particular to make this work.
The Technical Stuff – AMDs Bit
After an introduction, we were given a semi-technical overview of what AMD is doing as a processor manufacturer to integrate AI into its offerings. This overview mainly revolved around what is called “TOPS”, or “trillions of operations per second”, and the way this is achieved is by using a 3rd chip in the whole process.
To clarify, as many will know, commonly a computer had its main processor and a graphics processor (GPU). What has been introduced in the AI world is a 3rd called the NPU or Neural Processing Unit, and as I understand it, this handles the AI part of the stuff.
What makes AMD’s offering different is that cleverly, they have worked out a way of increasing the TOPS value – hence faster processing – from an initial figure of 16, to 20 and now to 50. AMD also claim at the same time, they have reduced the strain on the battery and are therefore quoting “all day battery life” in their spiel.
At this point, I hasten to add in terms of hardware, I am not a technical person – a claim also made by the spokesperson for AMD, Bryan Madden – so may have misinterpreted something in the translation, but I am sure anyone with a deeper knowledge than me will be fast to correct me, and please do in the Comments section below.
Other Speakers
Other speakers followed from HP and Microsoft discussing their respective company’s approach to AI, but we were also introduced to a filmmaker who I had never heard of, but apparently is quite famous in the YouTube world with a channel containing an alien fantasy series, Goldie Soetianto.
She described how she uses AI, and actually it turned out the Blackmagic Design DaVinci Resolve demo we saw earlier was one of hers as this was repeated. Next was a producer / filmmaker who had collaborated with Ms Soetianto to make a video showing her both at work and during the creation process.
It was explained that AI was used in the process of making this video.
Q & A
The final part of the evening was a panel discussion, followed by the bit that I was waiting for, the Q & A section. And it was in this, given the chance to sum up my thoughts, that I was offered the opportunity of the first question.
I prefaced it by saying I needed a few reference points. First was that while the DaVinci Resolve demonstration was indeed impressive, as a user of Resolve, I felt that to get to that level of expertise was certainly not for the beginner or even reasonably au fait user. There was LOTS of experience and a deep knowledge of Resolve and Fusion to get to there.
My second point was directed to the filmmaker / producer (who I apologise to as I misplaced the scrap of paper with his name, but as I can sort that, I will correct this story accordingly). I suggested that AI didn’t make the video, and indeed couldn’t. I asked if he agreed to create such a thing, required things AI just was not capable of doing, including vision, experience, skill, and importantly, emotion. He agreed, and actually added empathy to the list.
His further response was that AI tools assisted in the making of the video, making certain parts of the process faster.
My final point was that in the beta software demonstration (I think it was called either Aware or Amaze), a clip was shown of a very rough line drawing that a child made that was their drawing of a house. The software immediately turned this into something akin to an 18th century 3 storey mansion in full colour and 3D perspective.
Another clip showed AI removing a STOP sign from a photo.
My question, therefore, was considering the majority of the demos we had seen that day, allied with what most had described as the “pathetic” demonstration by Apple in previous weeks of its AI iteration, was it not so that AI was being put onto the mases with images of impossibly curled candy canes, purple mushrooms in generated cartoony forests and other silly pictures?
Additionally, was it not giving the impression that anyone using AI could immediately become an experienced architect or industrial designer? And finally, why were we demonstrating AI to show how things could be corrected, rather than teach people how to, for example, take a decent photo in the first place?
I finished with, as is my wont as people who know me understand, by saying it appeared that AI was being touted as the next big thing in the creativity world, but in my opinion, AI could never create something such as 2001: A Space Odyssey as it lacked the one major thing the human brain has, and that is emotion.
(It had only later struck me of the irony of using 2001 as an example considering AI, in the form of HAL the computer in the movie, tried to kill everyone off!)
In the main, the panel agreed with me on the creativity front, but went to great pains to also explain there were many other areas where AI was being developed such as the manufacturing, industrial design and medical disciplines.
The example was given my Bryan Madden from AMD of a medical scanner that could diagnose an ailment for a rural doctor when there was no local expertise in that field available, and send the results via the internet for confirmation and subsequent care / pharmaceutical suggestion by a qualified clinician in that field.
(Interestingly the writer of the original 2001: A Space Odyssey, Arthur C. Clarke, also foresaw this medical side of AI in the book, Rendezvous with Rama, when the on-board AI controlled robot surgeon made a mistake and killed the Mission Commander under surgery for a ruptured appendix).
In summation, for mine, and I stress I am only talking in the creativity area, the jury is still mainly out. There are some areas I am going to further look into such as the Resolve integration, and I also intend to delve deeper into Microsoft Co-Pilot.
As for imagery creation however, I still prefer to use my own imagination and skill to create MY vision of something, not a vision generated by, at the end of the day, a mathematical algorithm. And that goes for words to for that matter.
I can see the possibilities of AI tools assisting the creation process, and I hope this is where it goes as against taking that side from us (and thus possibly making the illustrators, animators, photographers, videographers and so on redundant).
Time will of course, tell. And as always, the development of AI, as history has shown time and time again, will follow the money.
Additionally …
Of course, a major part of the event design was to show off the new HP laptops with the AMD technology in them. And I have to say they are very impressive. On the flight back from Sydney, with my Gigabyte Aero over-spilling the not-so-large tray table, I wished I had one with me to be able to write this story at that time.
The HPs are fast, very fast, a great form factor and just the perfect size. For serious video editing, you’d want external monitors for sure, but the unit itself has more than enough processing grunt to take anything you’d throw at it. That part at least was ably demonstrated at the event.
As I say in the story, if you have thoughts on this subject, comments, want to tell me I am a raving lunatic or perhaps agree with me, let me know in the comments section, or email me at david@creativecontent.au
