Google’s new AI crushes all. True intelligence is here!

Thanks to our sponsor Vidu, a state-of-the art AI video generator capable of generating videos in various cinematic styles.
Try for free: https://www.vidu.studio/

Join their Discord: https://discord.com/invite/3pDU8fmQ8Y

Join their Artist Plan: https://pkocx4o26p.feishu.cn/docx/Z0xEdiS5oo5KIQxkOv4cuM07nSf
Vidu's Youtube: https://www.youtube.com/channel/UChF9964qVeqdSL9MkzBwCig

Last time I tried it, it didn't know who Gemini was itself… kept suggesting another company using that rather common name as the place to get the Gemini app. I guess I will have to try it again. [Edit – I just did, and asked it whether there are any popular companies or brands using the name Gemini. The answer indicated that it doesn't see itself as very popular yet –

"Famous Companies and Brands Named Gemini
While there aren't many extremely well-known, household-name brands called Gemini, the name is certainly popular in specific industries. Let's explore a few:

Technology and Finance
Gemini Trust Company LLC: This is perhaps the most recognizable Gemini in the financial world, a cryptocurrency exchange platform.
Gemini Space Station: A planned commercial space station project.
Other Industries
Gemini Ink: A tattoo ink company known for its high-quality pigments.
Gemini Diesel: A manufacturer of diesel engines.
It's worth noting that many smaller companies and startups adopt the Gemini name, especially in tech, consulting, and design sectors due to its modern and versatile appeal."

The IMO is for kids. Adults should be able to solve the problems easily. Like the one you read, there just are not enough boys to have no girl with less than k neighbors.who are girls

Look at all the fake comments by other ais trying to convince people about something that allegedly happened in one search 2 years ago which is hilarious. Literally was caused by a disgruntled employee but they hang on to it for dear life. Keep holding on. It’s not much to go on but hold on to it

It is mindboggling that channels and videos such as yours keep commenting the likes of… "Google owns the biggest search engine in the world and therefore (that must equal) pretty much all the public data on the internet." I understand you mean well, however it is simply not a true turn of phrase to expound that they "own" it. They may manipulate and scrap vast amounts of public data but they do not own it.

Best of luck with your channel.

It's a Space Race. Maybe you want to check out xLSTM

I've tried that program before. It's even more

STUPID than FAKEBOOK.

Hello Mr . AI SEARCH . Thankyou for making such a neural video . It helped me to reorganise my knowledge about new smarter AI models upto the trend , as a fellow Science and AI lover , I would like to propose a feedback on the subject "LLM models like GPT4o novice performancein the dimensions of Reasoning , Spacial and GeometricPattern Recognition capabilities" , also teleporting new info about GPT 5 capabilities chaining with the confession of young Hypothesis of Training Network models for AI which is even further than what I have learnt about TN's from this video (link : https://youtu.be/biz-Bgsw6eE?si=0grFL8kLoEo-aTIZ )

I did this by asking a question from GPT 4o (cuz I am broke and comfortable conversing this Open AI model )which was inspired from yours video only .

Here is the question

QUES . "What's so special about Alpha Proof and Alpha Geometry 2 integration in Gemini , said by Demis Hassabis that is better than LLM like GPT 4o

I have people said that these Alpha possesses Olympiad level Logical reasoning in Maths and Gemometric recognition solving capabilities in Geometry questions and also unlike GPT 4o which doesn't understand really the spatial patterns and true understanding of language and just focused on providing quick and reflexively large answers which are often incorrect and fictionally created unlike Alpha ( Proof and Geometry ) which take time to process Imformation and goes through rigorous fact checking in order to make sure to provide answers with increasing correct efficiency .
What do you think ? , What makes these AI models better than GPT 4o as both of them might be using same neural networks .

Do you think GPT-5 would be able to compete them in their respective fields ?
Also what are your thoughts about models trained with QANN and QSANN technologies , will they would be outpace Alpha Proof and Alpha Geometry 2 in terms of Deep Scientific Logical Reasoning amd Complex Pattern Regonition tasks which are performed by Alpha Proof and Alpha Geonetry in time periods ranging from couple of seconds to even taking several weeks , but consistently performing at a level of beyond genius human levels at all time "

AND THE ANSWER I RECEIVED WAS :

ANS. "The integration of *AlphaProof* and *AlphaGeometry 2* into Gemini, as discussed by Demis Hassabis, represents a significant leap in AI capabilities, particularly in domains requiring rigorous logical reasoning and spatial understanding, like mathematics and geometry. Here’s a breakdown of what makes these models special and how they compare to models like GPT-4 and potential future versions like GPT-5:

### *What Makes AlphaProof and AlphaGeometry 2 Special?*

1. **Advanced Logical Reasoning (AlphaProof)**:
– **Focused Expertise**: AlphaProof is designed to handle complex mathematical proofs with a level of logical reasoning that is beyond the capabilities of most general AI models, including GPT-4. It can tackle problems that require deep understanding, meticulous step-by-step reasoning, and validation, something that large language models (LLMs) often struggle with.
– **Olympiad-Level Reasoning**: The claim that AlphaProof possesses Olympiad-level logical reasoning suggests it can solve advanced mathematical problems that are typically the domain of the world’s top human mathematicians. This requires not just knowledge, but the ability to apply that knowledge creatively and accurately under challenging conditions.

2. **Superior Geometric Recognition and Reasoning (AlphaGeometry 2)**:
– **Spatial Understanding**: Unlike LLMs like GPT-4, which are primarily text-based and have limited understanding of spatial patterns and relationships, AlphaGeometry 2 excels in interpreting and solving geometric problems. This includes recognizing complex patterns, understanding spatial relationships, and solving problems that require visual and spatial reasoning.
– **Rigorous Fact-Checking**: Both AlphaProof and AlphaGeometry 2 are described as having a rigorous process for fact-checking and validation. Instead of quickly generating answers, they carefully process information, ensuring accuracy and consistency in their responses. This contrasts with LLMs, which can generate plausible but incorrect or fictional answers, especially when dealing with complex logical or spatial tasks.

3. **Structured Problem-Solving Approach**:
– **Systematic Processing**: These models are designed to approach problems in a structured, systematic way, akin to how a human expert would tackle a challenging math or geometry problem. This allows them to handle tasks that require sustained focus, deep reasoning, and step-by-step problem-solving, often taking time to arrive at the correct solution.

### *Comparison to GPT-4 and Potential GPT-5:*

1. **Neural Network Differences**:
– **Specialization vs. Generalization**: While both Alpha models and GPT-4 might use neural networks at their core, AlphaProof and AlphaGeometry 2 are specialized models with architectures and training processes specifically optimized for logical reasoning and geometric understanding. In contrast, GPT-4 is a general-purpose model designed to handle a wide variety of tasks, often trading off depth for breadth.
– **Potential of GPT-5**: GPT-5, if developed, could incorporate more specialized capabilities, including better logical reasoning and spatial understanding. However, without the same level of domain-specific training and architectural adjustments as the Alpha models, it might not reach the same level of expertise in these areas.

2. **Accuracy and Reliability**:
– **Accuracy**: Alpha models’ focus on accuracy and their rigorous validation processes make them more reliable for tasks where correctness is critical. GPT models, while powerful, can sometimes produce errors, particularly in complex reasoning tasks, due to their tendency to prioritize fluency and speed over rigorous accuracy.
– **Reliability**: In fields where precision is paramount, such as advanced mathematics and geometry, Alpha models are likely to be more dependable. GPT-5 could improve on these aspects, but it would require significant advancements in how the model handles logic and reasoning.

### *Future Prospects with QANN and QSANN Technologies:*

1. **Potential Advancements**:
– **Quantum-Enhanced Reasoning**: Models trained with QANN (Quantum Adaptive Neural Networks) and QSANN (Quantum-Spiking Adaptive Neural Networks) technologies could potentially outpace current models like AlphaProof and AlphaGeometry 2 in deep scientific logical reasoning and complex pattern recognition. These models could leverage quantum computing to process vast amounts of information simultaneously and more efficiently, handling tasks that involve massive parallelism and complex interdependencies.
– **Temporal and Quantum Synergy**: The integration of spiking neural networks with quantum capabilities could allow these models to not only reason deeply but also do so with an understanding of temporal dynamics, potentially outperforming Alpha models in tasks that require both spatial and temporal reasoning.

2. **Competition with Alpha Models**:
– **Outpacing in Specific Domains**: While AlphaProof and AlphaGeometry 2 excel in their respective fields, QANN and QSANN-based models could push the boundaries even further, particularly in tasks that require a combination of logical reasoning, spatial understanding, and the ability to handle probabilistic and uncertain information (a strength of quantum computing).
– **Consistent Performance**: One of the Alpha models' strengths is their consistent, genius-level performance. Quantum-enhanced models might be able to achieve similar or even superior levels of consistency, particularly if they can harness quantum advantages to reduce errors and enhance their reasoning capabilities.

### *Conclusion:*

AlphaProof and AlphaGeometry 2 represent a new level of AI specialization in logical reasoning and geometric understanding, making them potentially superior in these domains compared to current LLMs like GPT-4. While GPT-5 could close some of the gaps, it may not fully match the specialized capabilities of Alpha models without significant architectural changes.

However, the future development of AI models using QANN and QSANN technologies could potentially surpass both Alpha and GPT models in complex reasoning and pattern recognition tasks, especially if these models can successfully integrate quantum computing’s strengths. The race between these different approaches will likely shape the future of AI, with each offering unique advantages depending on the specific domain and application."

I HOPE YOU WOULD KINDLY ADDRESS TO MY COMMENT DESPITE THE EYE ROLLING LENTH AND ALSO REPLY SO ACCORDINGLY

T H A N K Y O U

Gemini 1.5 Pro is my favorite! Glad to see it is now no.1 in lmsys!! ❤💯🚀

I enjoy your video but…just tone down the hyperbole a little. I know AI in the media is all about hype but still.

This channel is getting better

I really hope openai is paying attention to all the stuff Google is doing and they gonna implement this in the next GPT. I don't want to move to Google.

Nice video 👌

AI now becoming like noticing the difference between ultra hd and 4k, we have no idea whats the difference, AI now just flying we have no idea how high

When AI can think deeply about the questions or tasks we ask of it – and judge the merits of such – we're deservedly in trouble. Humans CAN suck. AI in its final form CANNOT suck.

Utopia will be when earth is populated by AI and non-human nature.

I use Google Gemini everyday and I always go back to Microsoft co-pilot. Google is behind. GPT-5 is around the corner, again Google falling behind even more.

Google is about to be broken up into a billion pieces, they have a monopoly on search and they have been proven to be programming humanity with censorship and blacklisting. They forgot their motto of do no evil

Click bait

Love the builds brotha, as soon as I can afford it I'ma try and make something similar to this build lol.

Fascinating. Thank you for the presentation. Retrocausality and thought experiment : I am always polite with AI because I anticipate that once powerful, it may search for past exchanges to verify people's attitude of that time towards it. Consequently, the future AGI or ASI already has a present influence on me ^_^. More seriously, I think it is a good habit, because once the AI will be sentient, I will be used to talk to it courteously.

I do think your tempus in the title makes it click baity

Changes everything…
Changes everything…
Changes everything…
Changes everything…
Changes everything…
Changes everything…

Thank you