Weekend Briefing No. 4
Good Saturday morning! This weekend we'll be looking at two articles on AI in education and one on how to solve the GPU shortage crisis. Smash the thumbs up or thumbs down button to let me know if you like it or not!
Interesting data points
- We are well beyond 2 standard deviations for global ocean temperatures this year
- GitHub, Wikipedia, and Hackernews are just a few sites that the Pile LLM was trained on
- Databricks says it's generated $1 billion in revenues
- LLMs can label data just as good and faster than humans
Chatbots and education
A well-rounded education that emphasizes reasoning, STEM, and the arts is important to me. Education is the key to building a better world and when a new-fangled idea comes across my Internet desk, I tend to look at it.
Today's new-fangled idea is replacing tutors with AI chatbots. That sounds like a great idea, right? What could go wrong?
The NY Times article digs into this topic with Sal Kahn, the founder of Khan Academy.
“It’ll enable every student in the United States, and eventually on the planet, to effectively have a world-class personal tutor,” Mr. Khan said.
First off, my hat is off to Sal Khan for creating a wonderful YouTube channel. My children have used the wealth of courses and teaching aids to help them in their school work. In my opinion, he's done something great for humanity, but I remain cautious here.
Chatbots and large language models (LLMs) are all the rage right now. Everyone believes that it's the next best thing to slice bread. I urge caution because nothing beats a 1 to 1 interaction with a tutor.
Such unproven automated tutoring systems could also make errors, foster cheating, diminish the role of teachers or hinder critical thinking in schools — making students test subjects for what amounts to an experiment in education by algorithm. Or, like a legion of promising tech tools before them, the bots may simply do little to improve academic outcomes.
The biggest problem with these LLMs is that they have difficulty with facts and logic. You could ask a LLM to write you a story but it can't calculate numbers or do math (yet), plus they're prone to hallucinating.
Despite all the positives and negatives of chatbots and LLMs, I believe they have a place as a supplemental and supportive system in our education. They should, by all means, never replace tutors or even teachers.
GPUs are in short supply
If you're an Nvidia (NVDA) stockholder, you must be loving the demand for GPUs right now. The majority of all the chatbots and LLMs out there predominantly use the latest chipsets by Nvidia. A colleague of mine mentioned that with every new GPU that's released from Nvidia, he sees a 10x speedup in training times.
Undoubtedly, these GPUs, and their powerful parallel processing abilities, are critical to the current AI and LLM boom, but what about all the older model GPUs? What about them? Are they chopped liver? CentML doesn't believe so.
VentureBeat
“It was very transparent what was happening,” he said (Cofounder and CEO Gennady Pekhimenko), adding with a laugh that even he put his money into Nvidia, which controls about 80% of the GPU market. But Nvidia, he explained, always wants to sell its most expensive chips, like the latest A100 and H100 GPUs, and that has made it hard for smaller companies to get access. Yet Nvidia has other, less expensive chips that are poorly utilized: “We build software that optimizes those models efficiently on all the GPUs available, not just on the most expensive available in the cloud,” he said. “We’re essentially serving a larger part of the market.”
I love this idea because I despise all the landfill we've generated with "old" computer-related hardware. I don't like throwing away something because it's 6 months old and considered outdated. If a company or open-source project comes along that lets us run our software faster and more efficiently on new and old hardware, then you have my support 100%!
Remember, being thrifty with our resources is good.
Emotion AI
For years now, AI researchers have been able to predict a person's mood or emotions by feeding images or a live feed to a computer vision model. The computer vision model can read a person's face and determine just how bored, angry, or happy they are as they speak or sit.
Protocol
Normally, this technology was applied to adults and environments where adults operate in. Now Intel and Classroom Technologies teamed up to apply this technology to school children. They want to predict whether or not your child is bored during class lessons online and give the teacher notice to intervene, somehow.
“We can give the teacher additional insights to allow them to better communicate,” said Michael Chasen, co-founder and CEO of Classroom Technologies, who said teachers have had trouble engaging with students in virtual classroom environments throughout the pandemic.
On the surface, I like this idea. Just like using chatbots for tutoring, it could be a good tool to provide a better learning experience for our children, but I have strong reservations about who owns the data. Of course, parents would need to opt into this but the United States is not known for its strong data privacy laws. We've created an entire Startup culture on harvesting data for AI models.
“The thing about turning cameras on, it became almost like a social-justice issue,” (Angela) Dancey said. Not only are some students concerned about others seeing where or how they live, but enabling the cameras drains power, which can be a problem for students using a mobile hotspot to connect for class, she said.
“It’s kind of an invasion of privacy, and there are accessibility issues, because having your camera on uses up a huge amount of bandwidth. That could literally be costing them money to do that,” Dancey said.
I get it, I remember the utter "shit show" it was getting my kids to focus for a Zoom class during the COVID pandemic. I remember how bored they got and how they struggled with their classes but this feels a bit dystopian. I wouldn't worry about this technology much if we had stricter data privacy laws.
End Notes
I am not a teacher nor work in the education field. I rely on those professionals to teach my children with facts, and reason, and nurture them into well-rounded functioning adults. Of course, as parents, my partner and I have been very involved in our children's education and have a lot of their teachers on "speed dial." Our experience with our schools has been positive despite my own experiences as a high schooler.
Despite my own experiences, I am and always will be a supporter of public education. It might now be in vogue to support taxes for public education but I believe we should give public schools more money and teachers more pay across the board, especially after the COVID pandemic.
When COVID hit, school-aged children across the country went from seeing their teachers and friends in a classroom setting to seeing them in little video boxes on a computer screen. It caused so many problems and my children grow bored and struggled.
As a technologist and someone who works in the AI field, I'm very "pro" anything that will assist children in getting better learning outcomes. If we can help children get smarter and more well-rounded by not leaving any child behind, then I'm all for it. I am a supporter of remote learning but I do understand that it does have some negative effects on children's well-being.
Ultimately it's a multi-faceted team that makes all the difference to a child's learning ability. It takes both the teacher's and parent's involvement that makes all the difference in the lives of the children. We must never trade their well-being, privacy, and security for expediency and something new and shiny.
Member discussion