It isn’t surprising that Luma is asking for quotes on H100 inference capacity on LinkedIn today. Their recent release of Luma AI’s Dream Machine has stirred excitement in the tech world with its ability to create realistic videos from simple text prompts. I tested out the capability, but apparently, a lot of other people did as well. As someone intrigued by the creative potential of AI, I couldn’t help but write down a few thoughts about the latest offering of text-to-video in this space.
It is exciting to visualize your most creative ideas with just a few words. People who lack the deep technical knowledge required to create videos no longer need to invest in complex editing software or expensive equipment. Tools like Dream Machine make video creation accessible to anyone with a creative spark. This could lead to an explosion of content as more people explore these tools and flood the internet with AI-generated videos.
I know there are going to be many professionals who look at the output of these AI systems and criticize the quality of what is being created. They are far from something an experienced professional would create for a film or television product. Maybe that won’t ever be the target consumer, but quick and cost-effective video generation at scale could revolutionize entire industries. Advertisers could create highly targeted campaigns in minutes, educators could make abstract concepts come to life, and content creators could produce at an unprecedented pace.
The power of these tools is also missing some obvious safety messages. Every AI company, not just Luma, is pushing the technology as much as they can to remain competitive in a very lucrative market. There are still many serious ethical concerns about deepfakes, misinformation, and copyright infringement. We need to address these challenges as a society to harness the potential of AI video for good.
Competition in this area is also intensifying, with rivals like OpenAI and Kuaishou demonstrating impressive video generation capabilities. An open approach that encourages community engagement could give Luma AI an advantage, despite the fierce competition.
From a technical standpoint, creating coherent videos that adhere to prompts while maintaining natural movements is a significant achievement. Dream Machine has made progress, but there is still room for improvement, particularly in morphing effects and text rendering.
As a researcher and enthusiast in the world of AI, I’m fascinated by the possibilities of AI video generation. It’s a frontier full of potential that could transform how we create, but many deeper conversations also need to happen.
In the meantime, enjoy the waves, at least AI can create a relaxing scene while we try to figure out how this will impact us all.
Month: June 2024
The Professorless Post-secondary?
AI will revolutionize post-secondary programs and course creation. What does that mean for the future of universities and the value of human-led education?
To put this in context, I’ve designed decades worth of post-secondary programs ranging from 1-year certificates to 4-year degrees. The student value in creating these sizable learning experiences lies in their ability to provide a student with marketable skills in the job market and personal growth. For the business world, these programs provide the human capital they need to operate and compete successfully. Recently, I was reviewing a new course created by faculty, and it turned out to have been generated by AI.
It shouldn’t surprise me as I watch universities and professors complain about AI’s impact on the student work submitted to them. We should also consider AI’s implications for post-secondary business models. There are several key areas:
- Potential job displacement caused by automation: AI-powered platforms will be able to replace specific educational roles, such as course creation, assignment grading, and lecture delivery.
- Disruption of traditional teaching methods: AI-powered platforms will provide intelligent tutoring and adaptive learning, making education more accessible and personalized.
- Competition from online and alternative education providers: AI learning platforms will offer students more flexible, relevant and affordable options.
- Challenges adapting to AI and maintaining relevance: To remain competitive, post-secondaries will need to invest in expensive AI tools and platforms. They will also need to rethink the faculty role and the post-secondary experience, which may require significant changes to their business models and operations.
- Potential vulnerability to revenue streams: With other revenue streams and facing limits on tuition hikes, post-secondaries must look for new markets for revenue and students. International student recruitment has been a focus for many post-secondaries.
The timeline for disruption in post-secondary education is debatable, but given the pace of AI development, it will happen in the next 5-10 years. The end of post-secondary education isn’t inevitable, and they must develop strategies to adapt now. The strategies depend on the type of post-secondary and the learning experience they offer. Some common elements that all post-secondaries would need include:
- Ethical implementation of AI: Several ethical considerations, such as transparency, fairness, and data privacy, must be addressed in designing and deploying AI systems.
- Develop AI policies: Post-secondaries need to establish clear policies and frameworks to guide AI’s responsible and effective integration into teaching, learning and decision-making.
There are many other elements, but the key to this is constant evaluation and adaptation. One of the best descriptions I’ve seen of a successful future with AI is the concept of co-intelligence, which Ethan Mollick described in his book Co-intelligence. It is the idea that an individual (or institution) who understands AI’s abilities and limitations is uniquely positioned to realize AI’s full potential. A diversity of thought is beneficial, and post-secondaries will need to include the perspective of AI to reach the solutions and innovations required to navigate the disruption ahead.
References
Chan, C.K.Y. A comprehensive AI policy education framework for university teaching and learning. Int J Educ Technol High Educ 20, 38 (2023). https://doi.org/10.1186/s41239-023-00408-3
EULER University Institute. (2023, August 22). Artificial Intelligence (AI) as a threat to higher education – EFMU: The Euler-Franeker Memorial University and Institute. EFMU: The Euler-Franeker Memorial University and Institute. https://euler.euclid.int/artificial-intelligence-ai-as-a-threat-to-higher-education/
Graffiti Comet 