Google does not have an AI moat ?

Imagine a leaked internal memo from Google Researchers making some pretty bold claims. That’s what the AI world laid its hands on this week.

As always if you enjoy reading our posts be sure to spread the word !

Here’s what we have lined up for you today -

1. Google Researcher says the company does not have a moat, and neither does OpenAI

2. Fresh off the press

3. Meta’s new model has multiple senses

Google Researcher says the company does not have a moat, and neither does OpenAI

A researcher at Google has leaked an internal document that claims that Google (the company pouring billions into AI) does not have an edge when it comes to AI models and neither does OpenAI and for most part most LLM companies.

Instead the winner in this fierce AI race will be open-source models. Sounds crazy ? Probably is. Especially given that both OpenAI and Google have chosen to keep their newest generation of Large Language Models private.

The researcher argues that open-source has produced results that can compete with OpenAI and Google models at a fraction of the price. Ever since Facebook’s LLaMA model was leaked earlier in the year, individuals and organisations have been deploying open-source AI models that can -

1. Run on mobile phones with ease

2. Locally be hosted and on personal computers within an hour

3. Produce custom ChatGPT like results for under $100.

The researcher predicts that faster, more diverse and customisable open-source models are a matter of “when” and not “if” they will hit the market.

His prediction that much of the current value proposition provided by LLM companies will become irrelevant is bold and really somewhat of a stretch but what it implies is that there is a strong alternative emerging that users have a chance to access and potentially customise better.

The leaked memo is worth reading here !

Fresh off the press-

• The stock market rewards hardware companies that accelerate AI. AMD stock jumped a strong 10% on a report becoming public that the company is working with Microsoft to build chips

• Open AI has been burning cash to stay ahead as the market leader for LLMs. According to reports the company’s losses amount to $540 million and they could potentially try to raise as much as $100 billion over the coming years. That’s what they believe it will take to get to AGI (amongst many other things)

• Now that ChatGPT is an ecosystem of highly intelligent conversational bots, much of the product market fit seems to be taking shape (along with a business model of subscriptions, plugins and more). Sam Altman has shared that OpenAI is developing a pro-copyright model for ChatGPT. Yes, a model that will reward artists if their style and content is heavily repurposed for the model’s answers.

• Google I/O is starting in a few hours. Expect to see AI headlining multiple panels, and hopefully a lot of demos and public launches as well. We’ll most likely definitely see PaLM 2 (their SOTA large language model) and also how the company weaves in generative experience in their core search product

Meta’s new model has multiple senses

Image Bind by Meta. AI models with multimodal sensing capabilities all at once without requiring external supervision. Quite literally augmenting human senses in AI models.