GLOSSARY: Lang-O-Unchained

Can't read or see images? View this email in a browser

Borrowing brilliance from the open-source community, OpenAI recently repurposed a fan-favourite, LangChain, into a ‘function calling’ enhancement for its GPT-3.5-turbo and GPT-4 models, in its latest update.

This shiny upgrade to OpenAI API now empowers users to have dynamic and interactive conversations. But, the question is, has OpenAI stolen LangChain’s thunder? Let’s call this saga Lang-O-Unchained.

https://stratus.campaign-image.in/images/83238000108177016_9_1687348457231_ezgif-5-72d1ff969c.gif

“Yes, OpenAI took this from LangChainAI,” said Santiago, a machine learning expert. However, he believes that this will not kill LangChain. He said that there’s much more functionality there. “But how much more is OpenAI planning to lift from LangChain and integrate it natively on their API?” he pondered.

Only OpenAI will be able to answer that. However, we would not be surprised if the company starts integrating more features into its API, inspired by other frameworks and supports developed by LangChain and other open source initiatives.

LangChain was quick in responding with additional support. It is currently working on function calling support – converting existing tools to functions and developing new agents among other things. While the users are free to choose between the two, many developers believe that LangChain would remain relevant as it has other benefits. “The power of LangChain is that it rules them all, you can replace OpenAI with something else easily,” said Ivan Floravanti.

Read the complete story on the relevance of LangChain here.

Solving Imaginary Problems

There are two kinds of developers – those who tackle genuine problems head-on, and others who indulge in concocting imaginary issues, leading to the creation of flawed and unnecessary software. The latter poses a significant concern for developer teams and companies draining valuable resources.

But, in reality, hiring a developer who merely fulfils the assigned tasks without any hint of creativity is unlikely to impress companies. To truly stand out, developers are compelled to innovate and forge new paths. So, would you hire problem-solvers or those who dwell in the realm of the imaginary? Read to find out.

Indian IT & Generative AI

When OpenAI CEO Sam Altman said it was “hopeless” for India to build a general purpose LLM like the GPT models, Tech Mahindra CEO CP Gurnani tweeted: “Challenge Accepted”. Even though Altman’s comment was taken out of context and blown out of proportion, we will have to wait and see whether Tech Mahindra or other IT firms are building ChatGPT alternatives.

Unfortunately, there has been nothing substantial so far, besides R&D and partnerships. Here’s a quick glimpse of all the recent endeavours taken by Indian IT in generative AI.

https://stratus.campaign-image.in/images/83238000108177016_8_1687348456739_-1340x2048.png

Read the complete story here.

Lame LLM Papers

AIM recently put ChatGPT to the test by having it tackle India's toughest exam, UPSC, for the second time—it barely managed to pass. We noted that by altering the inquiry, we could eventually prompt the model to generate accurate responses. However, in the experiment, we considered only the first responses from the bot.

A recent paper by MIT researchers titled ‘Exploring the MIT Mathematics and EECS Curriculum Using Large Language Models’ revealed that GPT-4 achieved a perfect score of 100%, using a dataset of 4,550 questions and solutions. But, how? Read to find out more.

Welcoming Google 2.0

Google recently unveiled its new AI-powered Search Generative Experience at Google I/O 2023. A few weeks back it invited more people to try it out. Several beta users in the US have taken to social media to rave about the new upgrades. Here’s what they are saying.