Voice News

CA News 2024


A complete guide to 2024

Advantages Cons
The open model can be adapted to your business. Meta has trained Llama 3 on public information so that it, like other generative AI, can create harmful content or affect copyright law.
Generate text or images, especially with the Imagine feature for rapid image iteration. In our tests, the professional emails that Llama 3 generated were often a bit vague and fluffy compared to the emails written by ChatGPT using GPT-3.5.
Use with your existing cloud provider. As with other generative AI, training Llama 3 can take time and resources if you plan to modify the model.

OpenAI may be the better known name when it comes to commercial generative AI, but Meta has successfully carved out a niche for itself through open source powerful large language models. Meta unveiled its largest generative AI model to date, Llama 3, on April 18, which outperforms GPT-4 on some standard AI benchmark tests.

What is Lama 3?

Llama 3 is an LLM created by Meta. It can be used to create generative AI, including chatbots that can respond in natural language to a wide range of questions. The use cases on which Llama 3 was evaluated include brainstorming ideas, creative writing, coding, summarizing documents, and responding to questions in the voice of a specific persona or character.

The full Llama 3 model is available in four variants:

  • 8 billion parameters pre-trained.
  • 8 billion parameter instruction refined.
  • 70 billion parameters pre-trained.
  • 70 billion parameter instruction refined.

Llama 3’s generative AI capabilities can be used in a browser or through AI features in Meta’s Facebook, Instagram, WhatsApp and Messenger. The model itself can be downloaded from Meta or from major enterprise cloud platforms.

When will Llama 3 be released and on which platforms?

Llama 3 was released on April 18 on Google Cloud Vertex AI, IBM’s, and other major LLM hosting platforms. AWS followed suit, adding Llama 3 to Amazon Bedrock on April 23. From April 29, Llama 3 will be available on the following platforms:

  • Databricks.
  • Hugging face.
  • Kaggle.
  • Microsoft Azure.

Hardware platforms from AMD, AWS, Dell, Intel, NVIDIA and Qualcomm support Llama 3.

Is Llama 3 open source?

Llama 3 is open source, just like Meta’s other LLMs. Creating open source models has been a valuable differentiator for Meta.

SEE: Stanford’s AI Index Report Reveals 8 trends for AI in business Today. (TechRepublic)

There is some debate about how much of the code or weights of a large language model needs to be publicly available to count as open source. But in terms of business purposes, Meta offers a more open view of Llama 3 than its competitors do for their LLMs.

Is Lama 3 free?

Llama 3 is free as long as it is used under the terms of the license. The model can be downloaded directly from Meta or used within the various cloud hosting services mentioned above, although these services may incur fees.

The Meta AI homepage in a browser provides options for what Llama 3 should do.
The Meta AI homepage in a browser provides options for what Llama 3 should do. Image: Meta / Screenshot by Megan Crouse

Is Lama 3 multimodal?

Llama 3 is not multimodal, meaning it is unable to understand data from different modalities such as video, audio or text. Meta plans to make Llama 3 multimodal in the near future.

Improvements of Llama 3 over Llama 2

To make Llama 3 more capable than Llama 2, Meta added a new tokenizer to code language much more efficiently. Meta has supplemented Llama 3 with clustered query attention, a method to improve the efficiency of model inference. The Llama 3 training set is seven times the size of the training set used for Llama 2, Meta said, including four times as much code. Meta has applied new efficiency improvements to pre-training and fine-tuning instructions on Llama 3.

Because Llama 3 is designed as an open model, Meta has added guardrails with developers in mind. A new guardrail is Code Shield, which is intended to catch unsafe code that the model might produce.

What’s next for Lama 3?

Meta plans to:

  • Add multiple languages ​​to Llama 3.
  • Expand the context window.
  • Overall, increase the model’s capabilities in the future.

Meta is working on a 400B parameter model, which could help shape the next generation of Llama 3. In early testing, Llama 3 400B with instructional alignment scored 86.1 on the MMLU knowledge assessment (an AI benchmark test), Meta said, making it competitive with GPT-4. Llama 400B is said to be Meta’s largest LLM to date.

Llama 3’s place in the competitive generative AI landscape

Llama 3 competes directly with GPT-4 and GPT-3.5, Google’s Gemini and Gemma, Mistral AI’s Mistral 7B, Perplexity AI and other LLMs for individual or commercial use to build generative AI chatbots and other tools. About a week after Llama 3 was revealed, Snowflake debuted its own open enterprise AI with similar capabilities, called Snowflake Arctic.

The increasing performance demands of LLMs like Llama 3 are contributing to an arms race of AI-enabled PCs that can run models at least partially on-device. Meanwhile, generative AI companies may face increased scrutiny of heavy computing needs, which could contribute to worsening climate change.

Llama 3 vs GPT-4

Llama 3 outperforms OpenAI’s GPT-4 on HumanEval, a standard benchmark that compares the AI ​​model’s ability to generate code against code written by humans. Llama 3 70B scored 81.7, compared to the GPT-4 score of 67.

However, GPT-4 outperformed Llama 3 on the MMLU knowledge assessment with a score of 86.4 against Llama 3 70B’s 79.5. Llama 3’s performance on more tests can be found on Meta’s blog post.

Should your organization use Llama 3?

Assuming you’ve already decided that generative AI is right for your business, whether you want to use Llama 3 will likely depend on availability. Llama 3 is free to use and customizable within limits, more so than its competitors.

Llama 3 may be more effective than its rivals such as GPT-4 or Claude 3 for encryption. However, Llama 3 has other competitors in the coding field. For example, GitHub recently introduced Copilot Workspace, which is customized for coding and can create code from natural language prompts.

Llama 3 could be good for your organization if you want an open source family of general-purpose AI models.