November 4, 2023
Issue No. 77
CURATORS:
JOEL COLON
ASHRAF AHMAD
Welcome to this week's AtherXplorer, covering the latest in gaming, blockchain, and everything in between.
This week’s top picks include:
- AI: Stability AI Introduces New Tools including 3D and Image Fine-tuning
- Gaming: Disney’s ‘Dreamlight Valley’ Launch Announcement, Move from F2P to Paid Game
- Web3, Metaverse, and NFTs: OpenSea Lays off Half of its Staff, Prepares for “Opensea 2.0”
- Legal Landscape: Sam Bankman-Fried Guilty on all Charges
Artificial Intelligence
Stability AI Introduces New Tools including 3D and Image Fine-tuning
Stability AI has introduced new tools and previews:
- Sky Replacer, which allows users to easily replace the sky in photos with nine alternative skies. A useful tool for industries like real estate to showcase properties in ideal conditions. This empowers professionals across industries to enhance images efficiently.
- A private preview of Stable 3D was also announced, generating concept 3D objects from text or images within minutes without expertise. Generated in the standard .obj format, objects can be edited in tools like Blender and Unity, leveling the creative field for independent designers.
- Stable FineTuning provides enterprises and developers the ability to rapidly customize pictures into digital artworks like landscapes and avatars through turnkey integration. This is valuable for industries relying on visuals like entertainment, gaming, advertising, and marketing to personalize customer experiences.
- Furthermore, Stability AI has integrated Content Credentials and watermarking into its API to improve the transparency of generated images.
The company positions itself as delivering creative tools while focusing on open access through its heritage products and commitment to ethical, responsible AI development. Interested parties can apply to test upcoming offerings through the company's contact page.
Runway Releases Gen-2 Update, Further Improving Text2Vid and Img2Vid Generations
AI video startup Runway has updated its signature Generative AI video model Gen-2, impressing many users. Gen-2 now generates video from text or still images with significantly higher quality motion, resolution, fidelity, and consistency according to the company. It represents major advances in the software.
Originally launched in March 2023, Gen-2 improved on the prior Gen-1 model by generating 4-second videos from scratch or adding motion to uploaded stills. In August, it extended durations to 18 seconds and the following month added "Director Mode" tools to simulate camera movements like zooming, panning, and selectively adding motion.
The new update provides even smoother, sharper motion resulting in 2816x1536 resolution videos from still images, a notable boost from the prior 1792x1024. Artists can generate entire short productions by combining clips, with some stitching Gen-2 clips into music videos already screened in theaters.
Runway's founder views the software shift from manually pushing pixels to directing AI with language or parameters as redefining creativity. Early signs indicate AI filmmaking may emerge as a new creative force, mirroring how physical film grew in the 1920s.
The update coincides with ongoing Hollywood union negotiations over AI use, highlighting its potential to replace actors for short, silent works as Gen-2 does, at least temporarily, until creative partnerships are established. Many see the technology as advancing generative storytelling capabilities.
Microsoft Launches 365 Copilot for Enterprises
Microsoft has begun a quiet launch of its 365 Copilot AI assistant for enterprise customers, charging a $30 per month premium per user. However, access has limitations as businesses must commit to a minimum of 300 users and contact Microsoft directly to sign up.
Through Copilot, approved businesses can utilize AI-powered summarization, email generation, notes-based planning, and Excel analysis across Office documents. Microsoft hopes the premium offering will appeal to over 600 businesses that tested early access programs in recent months.
Access is currently restricted to only certain Microsoft 365 subscriptions like E3 and E5 commercial plans, excluding some Office 365 users. Monthly Enterprise Channel subscribers won't gain access until December.
While officially launching today, the rollout resembles a preorder as most lack immediate access. Copilot for Excel remains in preview as do the SharePoint integration starting in November and OneDrive integration in December.
Capabilities are also limited as Copilot for OneNote is Windows-only for now. This soft launch establishes Copilot's vision of transforming Office documents with AI but feels more like a first glimpse than full realization given pricing constraints and unfinished features still in testing across different applications. Forrester nonetheless predicts strong Copilot adoption among US knowledge workers by 2024.
Poe Introduces Creator Monetization for AI Chatbot Creators
Quora's AI chatbot platform Poe recently launched a program to financially reward those who create chatbots for its service. This establishes one of the first "creator economies" within the AI sector. Bot makers can earn money through two main ways - receiving a share of revenue if their bot leads a user to subscribe, or by setting a per-message fee that Poe will pay.
Originally introduced in February, Poe allows users to chat with bots from different companies like OpenAI, Amazon, and Google. The goal is for high-quality bot discussions to eventually spread from Poe to Quora's larger Q&A audience. Poe has gained users since launching, with over 18 million installs as of October according to market data. However, revenue remains low at around a quarter million per month currently.
The new monetization program is initially available only to US creators. Poe sets aside $10-20 from each subscription to split among bots that brought that user or showcased paywalls prior to their signup. This could help Poe compete against other chatbot apps making more revenue on mobile. Quora expects different types of useful bots to succeed like tutoring, media, entertainment, and more.
By financially rewarding bot builders, Quora aims to enable smaller groups to create bots accessed publicly despite limited resources. It also acts as an affiliate program of sorts to encourage more subscribers for Poe through creator-driven bots. This establishes one of the first creator economies within AI.
Brave Browser Releases New AI Assistant ‘Leo’
Brave has launched a new AI assistant called Leo, available for free to desktop users of its privacy-focused browser. Leo allows for translating, answering questions, summarizing webpages, and generating new content similarly to other chatbots. However, Brave asserts Leo provides stronger privacy as conversations are not recorded or used for model training, avoiding the privacy issues of competitors.
By default, Leo utilizes Meta's Llama 2 model but Brave is introducing a premium $15 monthly subscription featuring Anthropic's faster Claude Instant model as an alternative. The company stresses the importance of pairing AI with privacy given growing data concerns, positioning Leo as a privacy-first solution where users already spend time online in Brave.
Additional premium features include access to higher-quality conversations through prioritized queues and increased rate limits. Subscribers will also get early access to new capabilities while Brave indicates more models may be integrated over time, allowing users choice.
Leo was initially launched for Brave Desktop 1.60 and will come to Android and iOS in the coming months. Brave says it is committed to user privacy and providing secure, personalized assistance. The ability to substitute different models into Leo shows Brave's vision of users selecting among options for their AI interactions through the browser.
LinkedIn Unveils AI-Powered Premium Experience
LinkedIn has introduced new AI tools powered by OpenAI to provide personalized summaries and writing assistance to users. Initially available to Premium subscribers, the tools aim to enhance the LinkedIn experience and engagement. Users can apply the tools to their article feeds, write comments sharing content, or use them for professional correspondence like job applications.
LinkedIn is tapping OpenAI APIs through Microsoft Azure, combining OpenAI's large language models like GPT-4 with LinkedIn's proprietary user data. This personalized AI outputs differently for each user based on their profile and activities. Options allow framing summaries around questions to provide alternative perspectives.
The tools seek to help users more easily create and interact and overcome the challenges of solely professional networking where open self-expression can be limited. This aims to increase clicking, posting, and overall time on the platform. It also keeps LinkedIn relevant in an era where AI capabilities are increasingly expected.
While generative AI may initially attract through novelty, LinkedIn hopes it leads to valuable insights and networking. The expanded features additionally support LinkedIn's growing Premium business, which saw 55% year-over-year revenue growth and comprises an unspecified proportion of overall profits. Based on positive engagement, the company will evaluate expanding the tools to all users. Other media like video were also noted for future AI applications.