The video takes us to Meta's recent Kinect Developer Conference held at One Hacker Way, showcasing various technological advancements. Meta CEO Mark Zuckerberg highlighted new AI models - Llama 3.2 and its variants like Llama 3.21 b and 90 b, designed for developers and on-device use. These models, which incorporate vision and reasoning capabilities, support applications from business data analysis to imaginative features like visual grounding tasks. Meta emphasizes its commitment to open-source initiatives, presenting these models as competitive in the AI landscape while fostering a robust Ecosystem through cloud partnerships and safety tools like Llama Guard Vision.

In addition to AI innovations, Meta revealed its first true Augmented reality glasses called Orion, which employs Holographic displays and Neural tracking. These glasses, still in prototype form, are aimed at redefining everyday computing experiences while retaining a user-friendly design. Noteworthy is Meta's strategic integration of celebrity AI voices across its platforms and new interaction features, reinforcing the intersection of technology with entertainment. Meta's effort in AR reflects a vision of seamless digital integration into physical spaces, an approach setting a competitive standpoint against technology giants.

Main takeaways from the video:

Please remember to turn on the CC button to view the subtitles.

Key Vocabularies and Common Phrases:

1. Multimodal [ˌmʌltiˈmoʊdl] - (adj.) - Involving or using multiple methods, modes, or elements.

Vision models. Eleven B and 90 b were effectively Multimodal alternatives to Lama 3.18 b and 70 b.

2. Augmented reality [ɔːɡˈmɛntɪd riˈælɪti] - (n.) - A technology that superimposes digital information onto the user's view of the real world.

The big announcement wasn't about AI, it was about AR, Augmented reality, and specifically what Meta is calling its first true Augmented reality glasses.

3. Holographic [ˌhoʊl.əˈɡræf.ɪk] - (adj.) - Relating to the use or production of holograms, three-dimensional images formed by light.

Orion uses a pair of Holographic displays to place two d and three D content and experiences into your physical surroundings.

4. Benchmarking [ˈbentʃˌmɑːrkɪŋ] - (n.) - the process of comparing a company's products or processes against those of industry leaders to identify areas for improvement.

When it comes to Benchmarking, Meta presented llama 3.2 as competitive with leading foundation models for image recognition and visual reasoning tasks.

5. Neural tracking [ˈnʊərəl ˈtrækɪŋ] - (n.) - Monitoring or responding to neural activity, often used in enhancing human-computer interactions.

The glasses feature eye, hand, and Neural tracking.

6. Open source [ˌoʊpən ˈsɔrs] - (n.) - Denoting software for which the original source code is made freely available and may be redistributed and modified.

Part of the place that Meta has staked out in the AI Ecosystem is as the great defender of Open source.

7. Ecosystem [ˈiːkoʊˌsɪstəm] - (n.) - A complex network or interconnected system, especially in technology and business sectors.

The competition is not just at the state of the art, but to build a full Ecosystem of tools that are applicable for different cost contexts and different needs.

8. Commoditize [kəˈmɒdɪtaɪz] - (v.) - Turn something into a commodity, make its value based on its serviceability or satisfaction of needs.

Spending on models that it can then Commoditize forces the competition to lower prices.

9. Proliferation [prəˌlɪfəˈreɪʃən] - (n.) - Rapid increase in numbers or spread.

One of the profound benefits of Open source is the mass Proliferation of meta-related models.

10. Compute resources [kəmˈpjuːt rɪˈzɔːrsɪz] - (n.) - The computational power required to operate certain applications or models.

Building with them requires significant Compute resources and expertise.

Meta's AI and AR Revolution Unveiled at Kinect Conference

Today we are heading to One Hacker Way where Meta held its Kinect Developer Conference and unveiled a number of new things ranging from AI to AR. As you'll see, even on a day where we got a set of new models, it was really quite secondary to Zuckerberg's main presentation and a reminder that Meta has not entirely left the vision of the future that led them to rename the company Meta.

First, let's talk about the AI updates. Meta showed off the latest version of their open-source models at Llama 3.2. There was Llama 3.21 b and 3 b, designed specifically for on-device use cases, but for developers, the more standard were Llama 3.211 b and 90 b. Vision models, Llama 3.211 b and 90 b, were effectively Multimodal alternatives to Llama 3.18 b and 70 b. The multimodality is probably the biggest update, supporting image reasoning, such as document-level understanding, captioning of images, and visual grounding tasks, like pinpointing objects in images based on natural language descriptions.

For example, a small business owner could feed Llama 3.2 a chart of their last year of revenue and ask the model to highlight the months with the best sales. They could also show Llama a map and ask when a hike might become steeper. The models can be deployed with or without Meta's new safety tool called Llama Guard Vision, which can detect harmful texture images. The models are available to download from Llama directly, as well as from Hugging Face, and they're integrated into Meta's cloud partners.

These models are also being used to power AI features across Meta's social platforms. We didn't get an update to Meta's frontier level model 405 B, released back in August. Meta acknowledges that while these models are powerful, building with them requires significant Compute resources. They've also highlighted the competitiveness of Llama 3.2, presenting it as on par with leading foundation models for image recognition and visual reasoning tasks.

Meta's place in the AI Ecosystem is as a great defender of open-source. While most developers view Meta's commitment to Open source positively, there are skeptics. Meta's strategy with open-source tools might force competition to lower prices and spread Meta's AI widely. Meta is investing heavily in servers and infrastructure to train future models, underscoring a strategic business approach behind their open-source initiatives.

Meta isn't just thinking about tech specs, but also how they integrate these models into products. Meta showcases new use cases enabled by Llama 3.2 across platforms, while OpenAI has advanced their voice mode. Meta confirms voice interaction with Meta AI on Messenger, Facebook, WhatsApp, and Instagram DM, responding audibly. Celebrity AI voices are included, intersecting entertainment and technology. Meta demonstrates AI's image processing capabilities, identifying flowers or returning recipes from cake images. They're starting to test integrated translation on Instagram and Facebook, beginning with English to Spanish.

The focus was on consumers, but business use wasn't ignored. Click-to-message ads on WhatsApp and Messenger allow businesses to set up AI to interact with customers. There are over a million advertisers using these tools, resulting in numerous ads. Ad campaigns with Meta's Gen AI features saw higher interaction and conversion rates.

Community reactions are forming. There's positivity around open-source efforts. Dr. Jim Fan from Nvidia praises vision-language benchmarks for Llama 3.211 b, noting OSS AI models' strength globally. However, Meta's open-source models won't be in the EU. Meta also announced Orion, its first true Augmented reality glasses. Orion, similar to Apple Vision Pro, uses Holographic displays for immersive digital experiences in the physical world. They include eye, hand, and Neural tracking, with a wrist monitor feature that prevents public interaction awkwardness.

Orion is noted for its design, which resembles normal glasses. The prototype is not for sale yet, but Meta aims to polish it for consumer release. Some who tested it praised its impact, preferring Meta's careful development approach over rushing products. Observers noted the innovative pairing of wristbands with AR glasses, speculating it's a move similar to what Apple might have done.

The big takeaway from the conference was Meta's strengths in AI and AR, setting a course for innovative consumer interactions. Meta highlights the importance of robust developments over hasty releases, ensuring future technologies are impactful. Zuckerberg's leadership draws comparisons to Steve Jobs, hinting at Meta's potential industry transformation.

Artificial Intelligence, Technology, Innovation, Meta, Augmented Reality, Open Source