Exploring Gen-3 Alpha: Runway's Rival to OpenAI's Sora in 9 Captivating Videos

Welcome to Towards AGI, your premier newsletter dedicated to the world of Artificial Intelligence. Our mission is to guide you through the evolving realm of AI with a specific focus on Generative AI. Each issue is designed to enrich your understanding and spark your curiosity about the advancements and challenges shaping the future of AI.

Whether you're deeply embedded in the AI industry or just beginning to explore its vast potential, "Towards AGI" is crafted to provide you with comprehensive insights and discussions on the most pertinent topics. From groundbreaking research to ethical considerations, our newsletter is here to keep you at the forefront of AI innovation. Join our community of AI professionals, hobbyists, and academics as we pursue the ambitious path toward Artificial General Intelligence. Let’s embark on this journey together, exploring the rich landscape of AI through expert analysis, exclusive content, and engaging discussions.

TheGen.AI News

Exploring Gen-3 Alpha: Runway's Rival to OpenAI's Sora in 9 Captivating Videos

Artificial Intelligence is advancing rapidly, leading to both concerns about job displacement and excitement about its potential to enhance human creativity. U.S.-based company Runway AI has unveiled its latest AI model, Gen-3 Alpha, which they describe as 'a new frontier for high-fidelity, controllable video generation.'

Gen-3 Alpha is the inaugural model in an upcoming series trained by Runway on a new infrastructure designed for large-scale multimodal training. According to Runway, this model represents a significant leap in fidelity, consistency, and motion compared to its predecessor, Gen-2. This development marks a step towards creating General World Models, advanced AI systems capable of understanding the visual world and its dynamics.

Since the launch of Gen-3 Alpha, internet users have been showcasing their unique creations using the new model. These high-definition videos highlight the versatility and capabilities of Runway AI's latest offering. For instance, a user on X (formerly Twitter) known as Uncanny Harry AI created a short video of a fictional monster emerging from the Thames River in London. The cinematic 11-second clip features a 'hideous monster' rising from the river under a cloudy sky, reminiscent of Godzilla or Kaiju. Another user, Anu Akash, who explores AI tools, shared a time-lapse video generated by Gen-3 Alpha. The video depicts a pencil drawing of a girl with 'rabbit hair,' which Akash intended to be 'rabbit-like ears' but was pleased with the model's output despite the typo in her prompt.

Cognizant Boosts AI Suite with New Neuro Edge Platform

Global IT services provider Cognizant unveiled Cognizant Neuro Edge on Monday, adding a new tool to its Neuro suite designed to help businesses harness AI and generative AI.

Cognizant Neuro Edge is tailored for enterprise businesses, particularly manufacturers, integrating AI and GenAI with edge AI from chips and devices to applications, explained Vibha Rustagi, Cognizant’s global head of IoT and engineering. Cognizant, based in Teaneck, N.J., holds the No. 8 spot on CRN’s 2024 Solution Provider 500 list.

Rustagi described Cognizant Neuro Edge as a versatile framework that accelerates the development of enterprise edge services by simplifying model selection through quantization, fine-tuning AI and GenAI models, and using retrieval-augmented generation (RAG) among other techniques compatible with edge devices. This includes edge computing, which enables faster data access and real-time analysis.

The critical value of Cognizant Neuro Edge lies in providing businesses with hybrid GenAI cloud capabilities, Rustagi noted. It allows quicker data access, enabling businesses to utilize edge device compute resources more effectively. This leads to enhanced data privacy and security through localized computing, reduced cloud reliance, and improved efficiency with lower bandwidth and latency, thanks to real-time edge decision-making. Additionally, it offers advantages in personalization and contextual content, facilitating widespread GenAI adoption on the edge.

Neuro Edge is part of Cognizant’s broader Neuro platform, aimed at simplifying and speeding up the adoption of automation and AI in enterprises, specifically targeting edge applications, Rustagi added. It serves as an AI accelerator for devices and platforms, with a strong emphasis on edge devices like cars, medical equipment, surveillance cameras, and manufacturing machinery.

For instance, Neuro Edge can analyze data from specific manufacturing equipment, providing predictive maintenance and similar insights for device manufacturers or manufacturing plants, Rustagi explained.

While Cognizant has engaged with various industries, Qualcomm is the only named customer so far, according to Rustagi. She mentioned their collaboration with Qualcomm on automotive telematics and infotainment. Although not yet publicly announced, Cognizant is also working with other clients in industries such as med tech, manufacturing, and industrial IoT within the telecom sector. The partnership with Qualcomm involves deploying GenAI at the edge by integrating large language models (LLMs) within car infotainment and telematics systems, enhancing the driving experience through connected life services.

Cognizant Neuro Edge enables car manufacturers to offer real-time, personalized contextual awareness rather than generic information, unlocking new features in vehicles that were previously not possible, Rustagi concluded.

Illumex Raises $13M to Enhance GenAI for Structured Data

Large Language Models (LLMs) rely heavily on vast and diverse text data for training to generate human-like text. However, enterprises often face challenges in preparing their data for AI models despite understanding its importance.

Illumex Technologies Inc., a startup specializing in a Generative Semantic Fabric platform for structured enterprise data, leverages GenAI to prepare data for LLMs. The two-year-old company recently announced the successful closure of a $13 million seed funding round aimed at helping organizations tackle data challenges.

The funding round was led by Cardumen Capital, Samsung Ventures, and Amdocs Ventures, with additional contributions from ICI Fund, Iron Nation Fund, ICON Fund, Jibe Ventures, Ginossar Ventures, Today Ventures, and other notable angel investors.

Illumex aims to simplify the use of structured data for LLMs by automating the addition of a semantic layer. This layer unifies data silos and provides context to structured data, enhancing the reliability and relevance of AI interactions.

Inna Tokarev Sela, Founder and CEO of Illumex, emphasized the importance of resolving fundamental data issues for successful enterprise generative AI projects. She explained that their Generative Semantic Fabric aligns organizational data with business meaning and domain-specific context, enabling organizations to trust the results of their AI initiatives. This approach allows business users to interact with data using natural language without needing to learn precise technical definitions, thereby democratizing AI and empowering better decision-making.

A recent McKinsey study showed that 65% of enterprises are regularly using GenAI, a figure that has doubled from the previous year. Despite the value GenAI brings, concerns about its accuracy and reliability persist.

One significant issue with GenAI is the risk of hallucination. Some companies address this by integrating retrieval-augmented generation (RAG), which can enhance AI model performance. However, this method can be complex, particularly for structured data.

Challenges include the ongoing maintenance of data and the lack of semantic meaning in structured data. Manual labeling and contextualization of structured data can be resource-intensive and prone to human error. Illumex addresses these challenges by combining GenAI, graph databases, and relational databases to create a "Generative Semantic Fabric." This platform automates the discovery, mapping, and semantic enrichment of structured enterprise data, aligning it with business meaning and domain-specific context for reliable AI model outputs.

The platform also analyzes metadata associated with structured datasets without accessing underlying sensitive information, creating a unified semantic knowledge graph. This graph serves as a single source of truth, aligning all data with consistent business terminology and context.

Founded in 2021 by Inna Tokarev Sela, formerly AVP & Head of AI at Sisense and Senior Director of Machine Learning at SAP, Illumex has already partnered with several large enterprises, including Carson and Teva, despite operating in stealth mode.

Illumex has also collaborated with major tech companies like Microsoft, AWS, and Google Cloud, and offers integration with enterprise communication tools such as Google Meet, Slack, and Teams. The recent seed funding will help Illumex enable enterprises to derive greater value from their structured data, bridging the gap between AI innovation and stringent data governance and security requirements.

TheOpen.AI News

ITMO’s Open-Source AI Wins Gold at Russian Academy of Sciences Competition

ITMO-developed software clinched gold in the Artificial Intelligence and Quantum Technologies category at a competition for young scientists, organized by the Russian Academy of Sciences (RAS) and the Sistema Charitable Foundation. This innovative technology utilizes AI for enhanced data interpretation, applicable to a variety of scientific and industrial tasks such as flood forecasting and modeling data from technical sensors.

The software operates within an ecosystem of open-source tools designed to automatically create composite AI models using evolutionary optimization methods. These tools, developed by ITMO’s Research Center "Strong AI in Industry," have already been integrated into over 20 scientific and technological projects.

For example, organizations like the Arctic Science Center, Rosneft, Gazprom Neft, Center for Novel Technologies, and Poliinform use these technologies to predict natural and technological processes, from Arctic sea ice dynamics to detecting anomalies in pipeline magnetograms. The technology is also valuable in research, enabling fine-tuning of plasmatron arcs.

“At the competition, we showcased a technology that integrates these AI tools into composite models, rather than the AI tools themselves. Unlike other AutoML solution developers, we focused on model structure heterogeneity for maximum flexibility. Our team includes experts from various AI domains: Anna Kalyuzhnaya in probabilistic modeling and Alexander Khvatov in differential equations modeling. Another key to our success is our open-source approach, which helped build our ITMO.OpenSource community,” explained Nikolay Nikitin, the team leader and senior researcher at ITMO’s Laboratory for Composite AI.

This competition, dedicated to the 300th anniversary of the RAS, aims to support students and young scientists in developing and implementing innovative technologies and pioneering research in key sectors of the economy: chemical engineering, pharmaceutics, electronics, energy, and digital technologies. Organized by the Sistema Charitable Foundation and the RAS, with support from the Federal Service for Intellectual Property (Rospatent) and various Russian technology companies, the event recognized authors of 16 projects from 9 regions, selected from 347 projects submitted by participants from 45 regions. Winners included experts in hydrogen energy, breast cancer therapy, and hybrid protective coatings.

Open Source AI Coalition to Challenge Stable Diffusion with New Model

A newly established alliance of open-source AI developers is setting its sights on launching a new AI model to compete with Stable Diffusion and its latest controversial release.

The Open Model Initiative (OMI) was recently introduced by Invoke, a Generative AI platform for Professional Studios; Comfy.org, known for developing ComfyUI; and CivitAI, the largest Stable Diffusion repository globally. This group is leading a community-driven effort to create open-source AI models for generating images, videos, and audio.

“I aim for the OMI to release a next-gen model that the community can support and rally around within the next three to six months—an ambitious goal for a new organization, but one I believe we can achieve,” Kent Keirsey, CEO of Invoke AI, told Decrypt.

The OMI's objective is to develop open-source models that match or surpass the quality of proprietary models like Ideogram or MidJourney but without restrictive licensing terms. "Comfy is inclined to leverage a better-designed SD3," CivitAI noted in a recent post.

This initiative follows the release of SD3 by Stability AI, which faced criticism for its restrictive licensing terms. Despite the popularity of the Stable Diffusion family of AI models, the SD3 license was considered overly restrictive, leading to its ban from CivitAI.

To avoid such issues, the organization promotes truly open-source development and plans to adopt licenses like MIT or Apache-2, which ensure open, non-restrictive models with minimal conditions.

"We believe open source is the best way forward to ensure that AI benefits everyone,” the members stated in an open letter. “By collaborating, we can deliver high-quality, competitive models with open licenses that drive AI creativity forward, are free to use, and meet the needs of the community." The initiative also addresses ethical concerns by committing to develop a base model that does not include pre-trained capabilities such as "recognition of unconsented artist names" or "generating the likeness of unconsented individuals."

The initiative has already attracted significant support, with over 1,000 members joining its Discord server. "We’ve received more than 100 requests to join and support in just the last 24 hours," Keirsey told Decrypt.

Regarding funding, Keirsey mentioned that the initiative will not seek venture capital but will rely on community support and the business models of its founding members.

"The OMI will not accept investments, as the goal is to build open models, not generate profit," he said. "We have seen the negative impact when profit becomes the driving factor for organizations founded on the principle of open access to AI."

Each OMI member will maintain its business structure while participating in the group.

Robin Ken of ComfyUI explained to Decrypt that their role in “democratizing AI'' is at the tooling layer, unlike initiatives like Stability, Mistral, or Meta, which focus more on model development. He also confirmed that Comfy is financially backed and more than just a side project.

TheClosed.AI News

OpenAI's China Exit Opens Doors for AI Giants

This week, OpenAI has decisively restricted access to its site from mainland China and Hong Kong, effectively cutting off developers and companies in these regions from some of the most advanced AI technologies available.

OpenAI’s decision, though expected given the rising geopolitical tensions and tech rivalry, marks a significant moment in AI development that intensifies the ongoing tech cold war. This move will have substantial repercussions on the future AI landscape both in China and globally, setting the stage for fierce competition among AI superpowers.

Amid increasing governmental demands and the race for AI dominance, OpenAI’s action is a measure to protect its intellectual property while navigating complex geopolitical challenges. This move highlights the growing digital divide between China and Western countries, a key feature of the current tech war era. As OpenAI cuts ties with China, it signals a broader trend of tech decoupling, where the US and Chinese tech ecosystems are increasingly diverging, according to some experts.

Implications for Chinese AI Players

OpenAI’s blockade brings both challenges and opportunities for Chinese AI companies. Without access to OpenAI’s advanced models, such as GPT-4, the adoption and integration of cutting-edge AI technologies in China could slow, particularly impacting startups and smaller companies that lack the resources to develop similar models independently.

According to a report by the South China Morning Post, OpenAI’s move, effective from July 9, could affect Chinese companies that build services based on OpenAI’s large language models (LLMs). However, it could also spark innovation in China, pushing Chinese companies to develop their technologies, potentially leading to a new AI research boom and a more self-sufficient tech landscape.

The blockade also creates a gap that domestic giants like Alibaba, Baidu, and Tencent are well-positioned to fill. These companies have the financial resources, talent, and infrastructure to ramp up their AI research and development, fostering increased efforts in AI innovation and the creation of homegrown alternatives to OpenAI.

Furthermore, the Chinese government has heavily invested in its tech industry through large investments and favorable regulations. This could lead to a surge in AI research and competition among domestic players, aligning China more closely with its global counterparts.

Global AI Dynamics

OpenAI’s decision has global implications beyond China. This move could significantly shift the global AI landscape, potentially leading to a more fragmented environment. As the US and China vie for dominance, other countries and regions may align with one side based on their access to AI technologies.

Countries in Southeast Asia and Africa, which have strong economic ties with China, may favor Chinese AI solutions. In contrast, European and North American nations might increase their reliance on American-based AI technologies. This divide could impact international consortia, data exchanges, and the development of global AI standards.

The blockade also raises important ethical and security questions. OpenAI is exercising digital sovereignty, controlling who can benefit from its technology. This broader clampdown aims to ensure that AI technologies are built and deployed ethically and securely.

China faces a challenge in positioning its growing AI sector strategically to avoid appearing threatening to other nations. As the AI race intensifies, prioritizing ethics and international collaboration is crucial. Companies viewing China as an essential market will need to navigate complex geopolitical challenges.

For instance, Apple is reportedly seeking local partners to provide services that comply with Beijing’s stringent AI regulations, including standards set by the China Electronic Standardisation Institute last year. Ultimately, the future of AI depends not only on technological advancements but also on the geopolitical strategies and policies governing its development and deployment.

Open AI CEO Sam Altman Predicts GPT-5 to Surpass GPT-4's Flaws

There's been significant anticipation around OpenAI's launch of its highly anticipated GPT-5 model. Many expected it to be unveiled during the Spring Update event. However, the company introduced a new flagship model, GPT-4o, which OpenAI CEO Sam Altman described as "magical."

It appears that the wait for GPT-5 will be a bit longer. In a comprehensive interview at the Aspen Ideas Festival, Sam Altman discussed the development of "GPT-5" (via The Decoder). Although the launch date for GPT-5 remains uncertain, Altman claims it will be a "significant leap forward," especially compared to GPT-4. Previously, Altman expressed that GPT-4 "kind of sucks" and called it "mildly embarrassing at best."

Altman expects GPT-5 to be a major advancement, addressing many of GPT-4's shortcomings, such as its reasoning capabilities and tendency to make obvious mistakes. His comments suggest that GPT-5 might still be in the early development stages, likely due to complex algorithmic and data challenges, along with its scale.

The CEO likened the development of their LLMs to the iPhone's evolution. "The first iPhone was still pretty buggy, but it was good enough to be useful for people," Altman explained. OpenAI might be shifting away from traditional names and functions for its AI models, as indicated by a slide shown during ChatGPT voice demos at the VivaTech conference in Paris in May.

Altman's views on GPT-5 being a significant leap forward were echoed by a top executive at OpenAI, who predicted that today's ChatGPT would seem laughably inadequate within the next 12 months. We could be on the brink of a major technological breakthrough with AI. Altman noted that the AI revolution wouldn't require new devices, hinting that the technology might be cloud-based. However, most next-gen AI features for Windows 11 via the 24H2 release will need advanced Copilot+ PCs.

Altman also mentioned that "You'll be happy to have a new device" if the revolution necessitates new hardware. Global PC shipments are projected to grow by 8% by 2025, driven by the end of Windows 10 support and the rise of AI PCs. The economy is recovering post-COVID-19, leading to increased IT investments.

Additionally, a report indicates that OpenAI plans to release a new model later this year during the summer, which will reportedly make ChatGPT "really good, like materially better."

OpenAI Expands with Acquisition of Data Analytics Startup Rockset

OpenAI has acquired Rockset, an enterprise search and analytics startup, marking the most significant acquisition yet for the prominent artificial intelligence firm. On Friday, the maker of ChatGPT announced its plans to integrate Rockset’s technology to enable business customers and developers to better utilize their data on OpenAI’s platform. Additionally, some Rockset team members will join OpenAI. The financial terms of the deal were not disclosed.

OpenAI is in fierce competition with rivals like Anthropic and Alphabet Inc.'s Google to develop the most advanced AI models and bundle them with attractive services for businesses. The startup was valued at $86 billion in a tender offer in February and is expected to generate $3.4 billion in revenue this year, according to recent comments by CEO Sam Altman to employees. 

Rockset, founded by former Facebook employees, provides technology that allows businesses to retrieve, index, and make their data searchable. It serves industries ranging from gaming to fintech. In August, Rockset raised $44 million in a funding round led by Icon Ventures, with participation from Sequoia and Greylock, bringing its total funding to $105 million.

While OpenAI has previously engaged in acquihires—acquiring companies primarily for their talent—this acquisition is the first time it will integrate both the technology and the staff of the startup being purchased, according to an OpenAI spokesperson.

“Rockset’s infrastructure enables companies to turn their data into actionable insights,” said Brad Lightcap, OpenAI’s chief operating officer, in a blog post. “We’re excited to deliver these advantages to our customers by incorporating Rockset’s technology into OpenAI products.”

Imagine having Einstein's mind or Steve Jobs’ marketing skills in your team. You can do that now with AI. Join this FREE AI tools workshop to know how (Free for first 100) 🎁

In our quest to explore the dynamic and rapidly evolving field of Artificial Intelligence, this newsletter is your go-to source for the latest developments, breakthroughs, and discussions on Generative AI. Each edition brings you the most compelling news and insights from the forefront of Generative AI (GenAI), featuring cutting-edge research, transformative technologies, and the pioneering work of industry leaders.

Highlights from GenAI, OpenAI, and ClosedAI: Dive into the latest projects and innovations from the leading organizations behind some of the most advanced AI models in open-source, closed-sourced AI.

Stay Informed and Engaged: Whether you're a researcher, developer, entrepreneur, or enthusiast, "Towards AGI" aims to keep you informed and inspired. From technical deep-dives to ethical debates, our newsletter addresses the multifaceted aspects of AI development and its implications on society and industry.

Join us on this exciting journey as we navigate the complex landscape of artificial intelligence, moving steadily towards the realization of AGI. Stay tuned for exclusive interviews, expert opinions, and much more!

Keep reading