Google DeepMind's Gemini is Arriving Soon

Hey Everyone,

I think the culmination of Generative A.I. hype might occur when Google drops Gemini, the supposedly huge LLM that’s multi-modal that can compete directly with GPT-4, ChatGPT and other competitors.

From our Sponsor:

Designing Large Language Model Applications | Download the early release chapters of this O’Reilly generative AI ebook, compliments of Mission Cloud.

Have you downloaded the pre-release chapters from the latest O’Reilly ebook? Here’s what you’ll learn in chapter 3: LLM Interaction Paradigms, retrieval techniques, external tools, and much more. Start building valuable products that incorporate the power of language models.

Get your Copy -

Google Gemini is the first major Generative A.I. product by the newly merged Google brain and Deepmind teams now known as Google DeepMind. With a huge brain drain of talent involved in LLMs departing Google in recent years, many were disappointed with the capabilities of Google Bard, that seems to have been rushed, have limited utility and be simply a makeshift competitor to Bing AI.

Most expect Google Gemini to be a more direct competitor to OpenAI’s product, foundational model and Enterprise foundational model package with GPT-4. Google Gemini if it’s a success, would also make Google Cloud seem more appealing in the era of AI-as-a-Service.

Google is preparing to release a massive LLM called Gemini. With rumors suggesting it will launch this Fall or at the end of 2023. Most analysts believe that Google plans to make Gemini widely available through its Google Cloud platform, putting it in competition with OpenAI's API access.

Google's resources and experience in AI research and development may mean that it’s a more legit competitor to GPT-4 than anything we have seen before. If with less talent than OpenAI and perhaps even other A.I. startups, Google's training sources, include YouTube, Google Search, Google Books, and Google Scholar, may position Gemini as a strong contender it is believed by some.

While the likes of PaLM 2 seem impressive to me, I’m hoping Alphabet with its incredible sources and its founders back in the office, can make Generative A.I. (which they in a sense invented), a major force for Google’s continued innovation.

Google’s experience with PaLM and LaMDA is already considerable, so how will Google Gemini be different?

Take a Look at Google’s Ecosystem of AI

An Overview of Google’s AI Product Strategy

Read full story

Read: An Overview of Google’s AI Product Strategy

When it was reported that Google founder Sergey Brin was back at the office, after a divorce caused by a rumored affair with Elon Musk, you knew Gemini was going to be his baby.

Sergey’s net worth is $108 Billion, he’s not doing it for the money. He’s doing it for A.I. Supremacy. Gemini could be significantly more multi-modal that GPT-4 and will likely integrate with all the tools and AI products they have released and will continue to do so into 2024, and I’m talking dozens.

Sergey Mikhailovich Brin is an American billionaire business magnate best known for co-founding Google with Larry Page.

Gemini will also be a conversational A.I. system to rival ChatGPT. Let’s forget about Bing, Bard and all of that, Google managed to retain and even gain search advertising marketshare from Microsoft in 2022 and 2023.

We don’t know a ton about the technical details of Gemini yet, since it’s only been shared with select insiders and testers. Gemini as a GPU-rich entity, is also likely to have mechanisms to increase size and efficiency of their state-of-the-art large language models.

Sam Altman, who has 2 million followers on X, was cryping about the PR.

See the Tweet.

Silicon Valley Game of Thrones

The power-politics and heated histories between OpenAI, Elon Musk, Sergey Brin, Google is really getting personal. It’s a Silicon Valley game of thrones.

As for Google’s ability to use proprietary data, think about the potential use of YouTube video transcripts for training Gemini and other devious means. OpenAI has new lawsuits nearly every week.

The competition in the Large Language Model (LLM) space is intensifying, especially in the Enterprise space where OpenAI must compete with Cohere, Anthropic and a rising lists of pretty decent and well funded startups. Google for its part doesn’t want to be left behind, especially in comparison to its Cloud rivals, Microsoft Azure, and Amazon’s AWS.

In the marketing game of the race to AGI that OpenAI is on, Google could beat them to it. DeepMind’s ability to fine-tune LLMs to greater levels of efficiency could use cutting edge research and a huge budget to their advantage. While ChatGPT has its day, its consumer traction appears to be fading the further we get into 2023. Most users are young people, and besides side gigs and cheating on homework, young people prefer other Generative A.I. products like Midjourney, Character.AI

As a time when even YouTube Shorts have Generative AI upgrades and products, one has to wonder. The feature, called Dream Screen, will allow users to create an AI-generated video or image background just by typing in what you want to see. Sounds like Google is copying TikTok, with an editing app as well. How very original?

Meta will be releasing its AI personas as well (now it has in a disappointing event), which figures to be around 30. That should be announced around September 27th at one of Meta’s events. Meta hopes it will give users a new way to search, get recommendations, and otherwise engage with its products where you can now make multiple Facebook accounts, for different purposes.

Google’s Gemini is likely Alphabet playing defense, not offense, or was Bard defense and Gemini is offense? Meanwhile Google is in an anticompetition trial for the ages. It’s all quite the backdrop for Google Gemini to finally be released in November, December or January, possibly just weeks from now.

Google can also afford to make Gemini less expensive at roughly equal to the capabilities of GPT-4 and ChatGPT, which is clearly more expensive than many consumers are willing to pay for. A lot of OpenAI’s revenue likely come from its Enterprise products where in 2024 it will meet stiff competition.

GPT-4 is unlikely to be difficult for Google Gemini to replicate and improve upon, as mind-shattering as its performance was to some when it first came out. I think it was GPT-3.5 when I myself really took notice.

Google Gemini is Generative AI’s Act 2

While Sequoia wants us to believe that the Generative A.I. Unicorns are in fine form, outside of Midjourney and Charatacter.AI there haven’t been too many breakthroughs in 2023.

And if we are to believe the numbers by a16z, even a lack-lustre Bard is still being used by consumers.

See Tweet.

Google Gemini and its vast array of GAI products therefore has time to steal marketshare away from these over-funded AI unicorns ready to drift back into oblivion. Many verticals have far too many products for consumers to even appreciate.

In fact, Google Gemini’s real rival might actually be Inflection’s Pi, and not ChatGPT at all. ChatGPT for all its bloated virality, barely scratched the surface of mobile traffic even after a year.

A.I. Does Not Change Everything

Generative A.I. in fact changed 2023 dynamics very little. Even with all the luminaries and evangelists saying that “AI changes everything”.

For all intents and purposes in the bigger picture, all of these A.I. tools, gimmicks and product categories could amount to very little. They could be the “shitcoins” of Generative A.I. enthusiasm.

Generative A.I. is far more likely to be a convenient way for BigTech to sugar-daddy their moats. Google’s Gemini would I think be evidence of this. Who can compete with the GPU-rich, data and engineering of a team like Google DeepMind in 2023 who also have a Cloud presence with Google Cloud? Not even Amazon, Baidu or Meta can arguably compete with them. At least Nvidia has the sense to diversify their bets, Microsoft just bet on OpenAI and Inflection, for the most part.

This means that for Alphabet, the path to victory in Generative A.I. is still low-hanging fruit. Google Gemini thus has to deliver.

Meta can deliver an impressive open-source model to keep things interesting, OpenAI can get even more funds to create GPT-5, but the way PaLM, LaMDA and Gemini will improve means Alphabet definately has a somewhat easy path to victory. It could even fund Anthropic more and win from that angle as well. Google is highly likely to increase its share in Anthropic moving forward, since Anthropic is desperate to catch up with OpenAI.

Google thus has multiple paths to victory and they don’t have to even come through Gemini.

Google Will Likely dominate in B2B A.I.

Google could even own the “prosumer” as well in the B2B field of the future:

Google can keep iterating on Gemini, however disappointing or impressive it turns out to be. Even if OpenAI has better high-end talent and gets more funding. That’s not to say foundational enterprise LLM makers won’t be impressive and continue to grow their market cap and marketshare. Cohere, Anthropic, AI21 Labs and many others are impressive and closer to the capabilities of OpenAI than the media and AI bros give them credit for.

While some claim Generative A.I. is wide open, that’s not how the world works.

A world where BigTech have deeper pockets, better compute, more access to talent means a world where Generative A.I. is destined to be in the hands of the most powerful.

Whoever integrates Generative A.I. into their software services the best, wins. Google, Microsoft and Apple in a sense are in a league of their own. Generative A.I. will be used to leverage both Cloud marketshare and advertising performance. Eventually it will be making our videos for us. Google Gemini has the best chance of winning the multi-modal LLM arms race.

OpenAI’s first mover advantage will fade badly in 2024. Who will be left with serious momentum?

I think we can expect Google Gemini will be good, if not great. (as disappointing as Bard was).

While Google Brain lost too much talent in recent years, DeepMind is still decent, and that’s why they are now in charge. Sergey Brin needs to show the world that Google isn’t just some frauster that manipulates and fixes its own ‘Jedi Blue’ auction rigging with a legacy monopoly on search.

Google has a lot to improve, because we’ve mostly lost respect for it in 2023. Google search is a pretty terrible experience, hopefully Google Gemini will be something we can be proud of.

Will Google Gemini be an ace in the hole? Will it as an LLM have multi-modal and emergent properties not seen yet in other models?

Sundar Pichai CEO of Google recently noted: “While still early, we’re already seeing impressive multimodal capabilities not seen in prior models".

Thanks for reading!