GSV's AI News & Updates (08/08/23)
Instructure x Khan Academy, Meta's AI Chatbot Personalities, Einstein Studio, OverflowAI, Practical AI for Teachers and Students, Zoom's AI Policies, China's Robotaxis, The State of AI at Work
You are receiving this weekly update on the intersection of AI and education, as part of our GSV community – we hope this will be a helpful resource as we navigate this evolving and exciting intersection. Please feel free to send over any relevant resources you may come across (claire@gsv.com) – we’d love to learn and hear from you!
General 🚀
YouTube experiments with AI auto-generated video summaries: this initiative is part of YouTube's broader AI efforts, including AI quizzes for educational content and an AI dubbing tool.
Meta is reportedly preparing to release AI-powered chatbots with different personas: the chatbots are designed to have humanlike conversations with users on Meta’s social media platforms, including Facebook and Instagram. The report indicates that these chatbots will take on different personas, including one that advises users on travel plans in the style of a surfer and another that speaks like Abraham Lincoln.
Salesforce launches Einstein Studio: Einstein Studio is designed to make it easier for customers to use their own models to power Salesforce applications, without having to go through the process of building and training the models from scratch. It supports models built on Amazon SageMaker, Google Vertex AI, Databricks, and Snowflake.
ChatGPT launches Example Prompts and Suggested Replies: suggested prompts you can utilize at the start of a chat and recommended replies you can use in a conversation. The future of chat interfaces will be a combination of preloading context and then allowing AI to guide you to the information you seek.
Google Assistant to get an AI makeover: Amazon is making similar moves - the commerce giant is working on an AI-powered reboot for Alexa, its longtime digital assistant.
Stack Overflow announces OverflowAI: OverflowAI will initially focus on three areas: integrating generative AI into Stack Overflow's public platform, a Stack Overflow for Teams product, and an IDE integration. The company’s annual developer survey that revealed that the majority of developers want to use AI tools but only 40% actually trust AI.
Alibaba rolls out open-sourced AI models: marking the first time a major Chinese tech company has open-sourced its large language models. Companies with 100M+ MAUs will need a license from Alibaba to use the models.
Aided by A.I. Language Models, Google’s Robots Are Getting Smart: a sneak peek into Google’s new robotics model, RT-2, which melds artificial intelligence technology with robots.
SoftBank launches an OpenAI for Japan: SB Intuitions, building LLMs and generative AI in Japanese: the company, fully owned by SoftBank, will utilize data only from Japan-based data centers and will operate on a computing platform built on Nvidia GPUs, set to be online by March 2024.
a16z: The Economic Case for Generative AI and Foundation Models: explores the economics of traditional AI and why it’s typically been difficult to reach escape velocity for startups using AI as a core differentiator. BUT why generative AI applications and large foundation-model companies look very different, and what that may mean for our industry.
Education and the Future of Work 📚
Instructure and Khan Academy Announce Major Partnership On AI Tutoring, Teaching: this collaboration aims to combine Khan Academy's educational content with Instructure's extensive reach and data insights. The tool will also offer teachers insights into student engagement and performance, allowing them to tailor their teaching methods to individual students.
The state of AI at work in 2023: what customer service/sales employees really think.
Respondents aren’t scared about AI stealing their job. 70% of sales and customer service professionals say they don’t fear AI will steal their job one day.
Most companies don’t have an AI policy. 84% of respondents said their company does not have a company-wide AI policy.
Small businesses are most skeptical about using AI. 30% of companies with less than 50 employees are skeptical of AI tools, compared to 14% at companies with 51-100 employees, and 20% at companies with 100+ employees.
Those using AI are seeing the benefits. 79% of sales and customer service professionals using AI tools at work have experienced a positive impact on their performance at work.
GoStudent raises another $95M to go after VR and AI-enhanced tutoring: the company now says it has an “AI vision” and plans to “prioritize the creation of AI-driven tools”, such as lesson plan generators and AI tailored learning paths.
Ethan and Lilach Mollick’s Practical AI for Teachers and Students
Startups and Tools 🛠️
TextFX: Google Labs’ AI experiment that uses Google's PaLM 2 LLM. These 10 tools are designed to expand the writing process by generating creative possibilities with text and language.
Heuristica: AI-powered knowledge exploration, deep dives into any subject with AI-powered concept maps.
My AskiAI: your own WhatsApp AI chatbot, using your content in minutes. Save customer support from answering the same questions and drowning in documents, with faster, smarter answers.
GUS Universe — Website Builder: a designer in your pocket. Chat with GUS to lock in an idea. Sit back and watch the AI assistant work magic. In a matter of minutes, you've got a beautiful new site.
Tech 💻
LLaMA2 Chat 70B outperformed ChatGPT on the auto-evaluation leaderboard
Integration roundup: Monitoring your AI stack: how Datadog delivers end-to-end monitoring across every layer of your AI tech stack (from infrastructure to data storage to model serving to service chains and applications).
The Transformer Blueprint: A Holistic Guide to the Transformer Neural Network Architecture: a deep dive into Transformer a neural network architecture that was introduced in the famous paper “attention is all you need” in 2017, its applications, impacts, challenges and future directions.
Regulation and Policy ⚖️
The AI rules that US policymakers are considering, explained
OpenAI, Google, Microsoft and Anthropic form body to oversee safe ‘frontier AI’ development: new voluntary body strives to write its own rules before regulation takes shape.
Freedom of choice? How recent Zoom AI policy changes betrayed consumer trust: Zoom made changes to its TOS clarifying that the company can train AI on user data (meetings, private conversations, facial recognition), with no way to opt out.
Researchers Poke Holes in Safety Controls of ChatGPT and Other Chatbots: A new report by Carnegie Mellon University in Pittsburgh and the Center for A.I. Safety in San Francisco indicates that the guardrails for widely used chatbots can be thwarted, leading to an increasingly unpredictable environment for the technology. Report here.
ChatGPT tech to be adopted by Japan government for administrative tasks: Japan's new Digital Agency initially plans to spend just over $2 million to use the technology for one year, making it available to other ministries as well.
Other
Toyota and Pony.ai plan to mass produce robotaxis in China: the collaboration, which includes GAC-Toyota, will see a combined investment of over $1B yuan ($140M USD).
AI detects 20% more cases of breast cancer than human radiologists: AI has the potential to enhance the efforts of our radiologists, freeing up precious time for them.
How Silicon Valley is helping the Pentagon in the AI arms race
How can AI help with climate change?: are AI and ML helping in the climate fight, or hurting? Are they generating substantial greenhouse gas emissions on their own? Are they helping to discover and exploit more fossil fuels? Are they unlocking fantastic capabilities that might one day revolutionize climate models or the electricity grid?
Resources:
The GSV BIG 10: weekly coverage corner for the top 10 stories, insights, and major plays in learning and skilling.
Amazing as always. I was checking out the article on Instructure and Khan Academy joining powers. As a long-time user of products made by both companies, I am interested to see what is developed. But this quote caused me to be more than a little concerned:
“In Canvas,” Daly said, “teachers get great visibility into what’s happening. Teachers can say, ‘OK learning actually happened in this process.’ We have a lot of visibility on the engagement with content - how, when students engage content, information on how students are doing.” Daly continued, “when we start to bring all that information together with the Khan teaching content, we create a repository of information that teachers can use to modify their teaching delivery, student to student.”
“It is taking a cognitive burden off teachers, giving them time they can spend one-on-one with students – inspiring, mentoring and teaching, and creating more opportunity for collaboration, for in-person real engagement, student to teacher,” he said.
On the surface, this sounds like a win-win. But beneath the surface, these companies are getting access to the most detailed mental models of student learning in the history of education, and will absorb them into their data caches for undisclosed and unregulated uses. Think about it. Step-by-step reasoning printed out in a beautiful transcript; then plug the data back into the AI for further analysis. FERPA will struggle to keep up initially with these softwares. It will have to be modified to meet the challenges of these new technologies, but their most certainly will be an interregnum period. It will be interesting to watch how it all pans out. What we teachers have to think about in the coming days is not only how do we get the most accurate info about our students so we can make actionable changes to instruction but who ultimately controls the data?