Meta just rolled out automated budget scheduling and bid multipliers for its Advantage+ ad suite ahead of the holidays. Brands can now automatically bump up daily budgets during peak promotional spikes without rebuilding campaigns from scratch. It is also testing bid multipliers to automatically scale spend on high-priority audiences like Gen Z.
Mark Zuckerberg just laid out Meta's AI vision at a Senate forum, framing safety and access as the technology's two defining challenges. He defended open models like Llama 2 as crucial for American competitiveness and security, though Meta will rethink that stance if systems ever approach superintelligence.
Meta is developing a powerful new AI model designed to match OpenAI’s most capable system, according to a Wall Street Journal report. Expected to vastly outperform Llama 2, training is slated to start in early 2024 to help businesses build complex text generation and analysis tools. Meta has not commented.
NVIDIA just introduced TensorRT-LLM, an open-source library built to accelerate large language model inference on modern GPUs. It brings in-flight batching and FP8 quantization, delivering up to an 8x throughput boost on H100 GPUs and running Llama 2 4.6x faster than A100 systems. Why it matters: teams can now tap peak GPU parallelism via a modular Python API without writing low-level CUDA, slashing data center running costs.
Microsoft just warned that Chinese operatives are likely using AI-generated images to mimic U.S. voters and stoke division ahead of the 2024 election. Analysts say diffusion-powered visuals around hot-button topics like gun violence are far more convincing than old stock photos, helping covert operations spread virally to over 100 million people.
Meta chief AI scientist Yann LeCun explains why he rejects the existential risk hype around AI, arguing that open sourcing models like Llama 2 is essential to prevent a small group of Silicon Valley engineers from controlling the future interface of the digital world. For LeCun, broader access and transparency beat corporate gatekeeping every time.
Meta just re-licensed its flagship vision model DINOv2 under Apache 2.0, clearing it for commercial products alongside new dense prediction models for segmentation and depth estimation. The team also released FACET, a 32,000-image benchmark built to measure demographic bias in computer vision. Why it matters: developers can finally use DINOv2's visual features in commercial deployments.
The US Copyright Office just opened a public comment period to figure out where AI and copyright law collide. Regulators are zeroing in on three thorny questions: training on copyrighted data, granting copyright to AI-generated works without humans, and legal liability. Why it matters: As copyright lawsuits pile up against OpenAI and Midjourney, public input here will directly shape the playbook for future AI regulation.
Google just dropped a massive update to Vertex AI at its Cloud Next conference. PaLM 2 gets a 32k-token context window alongside quality boosts to Imagen and Codey, while third-party heavyweights like Claude 2 and Llama 2 join its Model Garden. With new extensions and enterprise search tools, Google is making a direct bid to outflank AWS Bedrock.
Meta is reportedly gearing up to train Llama 3 and Llama 4, with plans to bring Llama 3 up to GPT-4's performance while keeping it freely accessible under its open license. The detail surfaced from OpenAI researcher Jason Wei at a Meta event. If true, Meta is doubling down on commoditizing frontier models to rival closed labs.
Former Meta AI researchers just secured around $40 million to launch EvolutionaryScale, a new biotech startup valued at $200 million. The founding team previously developed a biological Transformer model predicting 700 million 3D protein structures, and they are now scaling up model sizes to tackle drug discovery and industrial biology.
Meta just dropped Code Llama, a specialized open-source coding model built on Llama 2 that scales up to 70B parameters. It handles code generation, debugging, and infilling across major languages, boasting a context window of up to 100k tokens. Why it matters: it gives developers a powerful, free alternative to proprietary coding assistants.
Meta just introduced SeamlessM4T, an all-in-one multimodal AI model that handles speech and text translations across nearly 100 languages. Instead of chaining separate tools, the unified system cuts errors and latency while supporting speech-to-speech translation into 36 languages. Meta is publicly releasing the model under a research license alongside 270,000 hours of multimodal alignment data.
Meta is bringing generative AI to daily chats, rolling out an experimental feature in WhatsApp's Android beta that turns brief text descriptions into custom stickers. Instead of hunting through sticker packs, testers can hit a new "Create" button and get tailored options. It's slated for a wider rollout in the coming weeks.
A report from Analytics India Magazine claims OpenAI spends around $700,000 every day to keep ChatGPT running, raising alarms about potential bankruptcy by late 2024. Despite heavy backing from Microsoft, shrinking web traffic and rising pressure from open-source alternatives like Llama 2 are straining its path to profitability.
Alibaba just open-sourced two 7-billion-parameter AI models, Qwen-7B and conversational variant Qwen-7B-Chat, supporting both English and Chinese. Global developers can build on them for free, while companies with over 100 million active users will need a royalty-free license. Why it matters: Alibaba is copying Meta's open-source playbook to win over developers and drive adoption of its cloud computing services.
A deep-dive analysis argues that generative AI and foundation models are breaking the historical mediocrity spiral of legacy software. Unlike older systems crippled by long-tail accuracy costs and human-in-the-loop overhead, consumer hits like ChatGPT and Character.AI are unlocking unprecedented organic adoption. Why it matters: Startups finally have the growth velocity and defensibility to challenge entrenched incumbents rather than ending up as bespoke IT services.
Meta just open-sourced AudioCraft, a unified codebase designed to generate high-fidelity audio and music from simple text prompts. The release packages together MusicGen, the sound-effect generator AudioGen, and an improved EnCodec neural decoder, complete with open weights and code for researchers to train their own models.
Instagram is working on notices that flag when content was "created or edited with AI," according to app screenshots uncovered by a researcher. The move follows Meta's recent White House pledge to develop watermarking systems, suggesting the platform may proactively label synthetic images instead of relying purely on user disclosures.
Meta is preparing to launch AI chatbots with distinct personalities across Instagram and Facebook as soon as next month. Powered by its LLaMA models, the characters reportedly range from a Lincoln impersonator to a chill surfer offering travel advice. Why it matters: The interactive characters aim to fend off TikTok, prove Meta's generative AI chops, and capture rich user interest data for ad targeting.
