Meta Introduces New AI Language Model Called LLaMA That’s Similar To Chatbots
The race to produce the best chatbot in the industry is on as Facebook’s parent firm makes a shocking new unveiling.
The
company has launched its own AI language model called LLaMA and it
works in a similar manner to the viral chatbots taking the lead today.
Moreover, the firm says that it even has the capability to outperform
others in the industry like ChatGPT by OpenAI.
The tool has the
capability to function in a manner that’s awfully similar to a regular
chatbot. This means it can generate human-like chats in a manner by
which other AI chatbots function. But at the same time, Meta boasts how
it runs in a more efficient manner than other models taking the lead.
Also, they claim it features much fewer requirements linked to hardware.
One
of the tech giant’s researchers mentioned how LLaMA may be around ten
times smaller in size but it has the capability to outperform in so many
different ways. However, despite all these wonderful claims, it’s only
designed for release to researchers.
They wish to gather more
insights on the product from them, including constructive criticism that
would help it become the best tool out there. Only then would Meta like
to allow for these chatbots to come forward and provide the masses with
benefits.
Meta similarly hopes to eliminate the possibility of
any biases and factual mistakes that are commonly prevalent in tools
like ChatGPT. And the benefit of running on just one GPU means it can
further enhance access to language models.
By the looks of it so
far, things definitely do appear to be promising. But we must remember
that in the past, Meta’s track record for such chatbots isn’t too great.
They failed to generate the same sort of excitement and
popularity that we’re seeing for ChatGPT and the new AI Bing search by
Microsoft. Nearly 100 million are making use of ChatGPT and there are no
signs of them stopping just yet.
Hence, this might be one
reason why it really wishes to take out some time before a launch to
ensure all things are as perfect as can be.
On the other hand,
Facebook’s parent firm says that attaining access to big language models
is not always easy and most people are faced with great limitations.
They need servers for functioning and access is often restricted in
terms of understanding how such models work. The end result is a big
roadblock in terms of its robustness and a more likely chance of
producing misinformation.
Meta did detail how its AI language
model arises in four different variants and it operates in a variety of
different parameters. This is much smaller than what we’re seeing for
the currently popular ChatGPT.
Meta claims LLaMA could outperform
the likes of bigger language models by enabling it to train on some
more data snippets which it called tokens.
So far, they’ve conducted their training well on trillion tokens and text from the web.
