Llama 2 from Meta signals worry for OpenAI’s ChatGPT: What is it & how to access

Meta said that Llama 2 will unlock more progress across the AI ecosystem. The company has launched pretrained and fine-tuned models with 7B, 13B, and 70B parameters. In his post, Zuckerberg said that Llama 2 was pre-trained on 40 per cent more data than its predecessor and also has improvements to its architecture. For now, users can download these models through Azure along with Microsoft’s safety and content tools. 

Llama 2 decoded: 

Llama 2 was created within a research lab in Meta, by a team led by Yann LeCun who is hailed as one of the founding fathers of neural networks. Reportedly, Llama 2 has outperformed GPT-3 and is similar to ChatGPT when it comes to chat model quality. However, it lacks the ability to code.

The company has carried out some extensive safety evaluations and has deployed techniques such as context distillation and Reinforcement Learning from Human Feedback (RLHF). Context distillation is a method to internalise information from abstract instructions, while RLHF is a technique that trains a reward model based on human feedback.

Llama 2 will be available for commercial use when a product made using the model has over 700 million monthly active users. While it is available via Microsoft’s Azure platform, AWS, Hugging Face; Qualcomm is collaborating with Microsoft to integrate the Llama 2 model into phones, laptops, and headsets from 2024. Considering the breadth of potential innovations that Llama 2 has, AI experts feel that it can be a disruptor in the industry. 

What can Llama 2 do?

Llama 2 is a large language model (LLM) that can be used to create chatbots such as ChatGPT. LLMs are trained on humongous amounts of data that allow them to replicate human language and even programming. Now that Meta is making Llama 2 open-source, its content will be available for people to access for free and use and tweak to create their own AI chatbots for other purposes.  

The objective behind making Llama 2 open source is so that start-ups and other businesses can access it and tweak it to create their own AI products. This could potentially add to the ever-increasing list of chatbots that can rival OpenAI’s ChatGPT or Google Bard. 

When asked why Meta is releasing Llama 2 for public use, Meta’s president of global affairs, Nick Clegg told BBC Radio 4 that making LLMs open-source would make them ‘safe and better by inviting outside scrutiny’.

It is also believed that with the open-sourcing of Llama 2, Meta is trying to weaken the competitive edge of OpenAI, Google, etc. Interestingly, Microsoft is the biggest investor in OpenAI, yet it is backing Llama 2. 

How can you try Llama 2?

As of now the easiest way to try Llama 2 is via Perplexity AI. Visit llama.perplexity.ai to explore the model. The link will open a chatbox with the greeting “Hello! I am a LLaMa created by Meta AI and brought to life by the Perplexity team. How can I help you?” Users can enter their queries in the input box at the bottom. 

We asked the Llama model through Perplexity AI to write a 100-word essay on British colonialism and below is the answer. 

“Sure! Here is a 100-word essay on British colonialism: 

British colonialism had far-reaching impacts across the globe. From the Americas to Africa, Asia, and Oceania, the British Empire expanded its reach through conquest, trade, and colonisation. While it brought modern infrastructure and technologies to many regions, it also imposed its own culture, language, and values upon indigenous populations. The legacy of British colonialism can be seen in the political, economic, and social structures of former colonies today. However, it is important to acknowledge both the benefits and drawbacks of this period in history, recognizing the suffering and exploitation experienced by many communities under colonial rule.”

Concerns surrounding open source for AI?

There has been widespread apprehension in the tech community about open-source – AI. The primary concern is that open-sourcing would likely make powerful AI tools accessible to all, including bad actors like hackers and criminals. As of now, Meta requires users to agree to an acceptable use policy which includes not using LLMs for disinformation or encouraging violence or terrorism. The Llama 2 model comes with a responsible use guide for developers.

Related Stories

IndiGo employee alleges caste-based abuse by senior officials; airline ‘strongly refutes’ claims

Test Priyanshu Haldar

‘Attack on Constitution’: After BJP MPs’ comments on judiciary, Congress senses opportunity, plans to corner government

Why so many Indian students — especially from Kashmir — go to Iran for higher education