The Unlikely Roots of Large Language Models: U.S. military funding helped build the foundation for ChatGPT and other innovations in natural language processing.
Technical Insights

The Unlikely Roots of Large Language Models: U.S. military funding helped build the foundation for ChatGPT and other innovations in natural language processing.

I’d like to share a part of the origin story of large language models that isn’t widely known. A lot of early work in natural language processing (NLP) was funded by U.S. military intelligence agencies that needed machine translation and speech recognition capabilities.
The Hidden Value of Deep Technical Knowledge: What's the best way to prompt a large language model? Understand the technology.
Technical Insights

The Hidden Value of Deep Technical Knowledge: What's the best way to prompt a large language model? Understand the technology.

Machine learning development is an empirical process. It’s hard to know in advance the result of a hyperparameter choice, dataset, or prompt to a large language model (LLM).
Tips for Taking Advantage of Open Large Language Models: Prompting? Few-Shot? Fine-Tuning? Pretraining from scratch? Open LLMs mean more options for developers.
Technical Insights

Tips for Taking Advantage of Open Large Language Models: Prompting? Few-Shot? Fine-Tuning? Pretraining from scratch? Open LLMs mean more options for developers.

An increasing variety of large language models (LLMs) are open source, or close to it. The proliferation of models with relatively permissive licenses gives developers more options for building applications.
Does AI Understand the World?: There's no scientific test for understanding, but there is evidence that large language models understand the world to some extent.
Technical Insights

Does AI Understand the World?: There's no scientific test for understanding, but there is evidence that large language models understand the world to some extent.

Do large language models understand the world? As a scientist and engineer, I’ve avoided asking whether an AI system “understands” anything. There’s no widely agreed-upon, scientific test for whether a system really understands — as opposed to appearing to understand —
Building Machine Learning Systems is More Debugging Than Development: Traditional software development requires meticulous planning. With machine learning, it's better to jump right in.
Technical Insights

Building Machine Learning Systems is More Debugging Than Development: Traditional software development requires meticulous planning. With machine learning, it's better to jump right in.

Internalizing this mental framework has made me a more efficient machine learning engineer: Most of the work of building a machine learning system is debugging rather than development.
AI at the Speed of Prompting: Prompt-based development enables you to try out ideas quickly and cheaply — no need to scope projects carefully.
Technical Insights

AI at the Speed of Prompting: Prompt-based development enables you to try out ideas quickly and cheaply — no need to scope projects carefully.

Prompt-based development is making the machine learning development cycle much faster: Projects that used to take months now may take days. I wrote in an earlier letter that this rapid development is causing developers to do away with test sets.
Breakthroughs on the Horizon?: Innovations in computer vision stole the spotlight at this year's CVPR conference.
Technical Insights

Breakthroughs on the Horizon?: Innovations in computer vision stole the spotlight at this year's CVPR conference.

I spent Sunday through Tuesday at the CVPR computer vision conference in Vancouver, Canada, along with over 4,000 other attendees. With the easing of the pandemic, it’s fantastic that large conferences are being held in person again!
Existential Risk? I Don't Get It!: Prominent computer scientists fear that AI could trigger human extinction. It's time to have a real conversation about the realistic risks.
Technical Insights

Existential Risk? I Don't Get It!: Prominent computer scientists fear that AI could trigger human extinction. It's time to have a real conversation about the realistic risks.

Last week, safe.org asserted that “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.
Building AI Systems No Longer Requires Much Data: Pretrained models make it possible to build AI systems using very little additional data.
Technical Insights

Building AI Systems No Longer Requires Much Data: Pretrained models make it possible to build AI systems using very little additional data.

It’s time to move beyond the stereotype that machine learning systems need a lot of data. While having more data is helpful, large pretrained models make it practical to build viable systems using a very small labeled training set — perhaps just a handful of examples specific to your application.
Beyond Test Sets: How prompting is changing machine learning development
Technical Insights

Beyond Test Sets: How prompting is changing machine learning development

A few weeks ago, I wrote about my team at Landing AI’s work on visual prompting. With the speed of building machine learning applications through text prompting and visual prompting, I’m...
"Visual Prompting” Builds Vision Models in Seconds: A new approach applies ideas from text prompting to computer vision
Technical Insights

"Visual Prompting” Builds Vision Models in Seconds: A new approach applies ideas from text prompting to computer vision

My team at Landing AI just announced a new tool for quickly building computer vision models, using a technique we call Visual Prompting. It’s a lot of fun! I invite you to try it.
Figure showing how researchers obtained the Alpaca model
Technical Insights

When One Machine Learning Model Learns From Another: Was Google’s Bard trained on output from OpenAI's ChatGPT? The technique is legit, but it raises thorny questions.

Last week, the tech news site The Information reported an internal controversy at Google. Engineers were concerned that Google’s Bard large language model was trained in part on output from OpenAI’s ChatGPT, which would have violated OpenAI’s terms of use.
Emad Mostaque, Alexandr Wang, Andrew Ng, and Peter Diamandis at Abundance 360, March 20, 2023
Technical Insights

Catching AI's Next Wave: Generative AI will drive tremendous value and growth.

Generative AI is taking off, and along with it excitement and hype about the technology’s potential. I encourage you to think of it as a general-purpose technology (GPT, not to be confused with the other GPT: generative pretrained transformer).
IBM Watson wins the television game show Jeopardy!, February 16, 2011
Technical Insights

AGI Progress Report: The latest AI models are exciting, but they're far from artificial general intelligence

Here’s a quiz for you. Which company said this? “It’s always been a challenge to create computers that can actually communicate with and operate at anything like the level of a human mind...
Landing AI's computer vision platform, LandingLens
Technical Insights

Computer Vision Made Easy!: LandingLens enables anyone to build in minutes models that used to take months.

Landing AI, a sister company of DeepLearning.AI, just released its computer vision platform, LandingLens, for everyone to start using for free. You can try it now.
Load More

Subscribe to The Batch

Stay updated with weekly AI News and Insights delivered to your inbox