Google has announced a new AI system called Bard that can generate images and text responses based on conversational prompts. Bard signifies Google’s major investment into large language models and shows its commitment to being an AI leader after the viral success of ChatGPT.
Background on Large Language Models
Over the past few years, there has been rapid progress in developing large language models – AI systems trained on huge datasets of text data. Models like GPT-3 and Google’s PaLM can generate human-like text and power applications like chatbots.
More recently, startups like Anthropic and research labs like DeepMind have produced models that can not only generate text, but also create images and other multimedia. The viral interest in these creative AIs highlights their potential.
Google has been lagging behind in this AI wave, with most of the hype going towards ChatGPT and tools from its competitor OpenAI. Bard represents Google’s big entry into the conversational AI space to take on these rivals.
Introducing Bard and its Image Capabilities
Sundar Pichai officially announced Bard on Monday February 6th, describing it as an “experimental conversational AI service” backed by Google’s language research and compute. The name “Bard” ties into the idea of storytelling and creativity.
Unlike ChatGPT which is exclusively text-based, Bard can generate images to accompany its written responses. This allows richer, multi-modal conversations spanning text and visuals.
Some examples that Pichai provided include asking Bard to:
- “Explain recent discoveries from the James Webb Space Telescope in a poem” – Bard provides a poem summarizing cosmic images of star nurseries
- “Create a short video about hiking trails in the Swiss alps” – Bard produces a 1-minute video highlighting Alp hikes
This combination of text and synthetic media generation makes Bard stand out from previous conversational AI models.
Behind the Scenes – LaMDA and PaLM
Bard is built on top of Google’s existing natural language AI systems – most notably LaMDA and PaLM.
LaMDA (Language Model for Dialogue Applications) is a text-based conversational model capable of human-like discussions on various topics. Google has been testing LaMDA since 2021 in research experiments.
PaLM (Pathways Language Model) is Google’s enormous general purpose language model announced last year. PaLM has state-of-the-art performance on language and text tasks thanks to its 540 billion parameters.
Bard effectively combines these models, using LaMDA’s conversational abilities and PaLM’s broad language mastery. The huge compute available in Google Cloud also enables complex image generation.
Gradual Public Launch to Rival ChatGPT
Google plans a slow rollout of Bard to gather user feedback and avoid problems that plagued competitors like Microsoft’s viral Bing chatbot.
Initially, access will be limited to trusted testers and enterprise clients via Google’s cloud platform. This contrasts ChatGPT which received widespread consumer access immediately after launching.
Over time Bard will integrate with Google Search and become available for more users. The goal is for Bard conversations to provide richer, more in-depth explanations to search queries.
Google wants Bard to meet a high accuracy bar before permitting universal access. Models will keep training on new data to improve capabilities and trustworthiness.
Racing to Lead the AI Industry
Launching Bard signifies Google entering the arms race for supremacy in generative AI:
| Company | Key Models | Focus | Stage
| ———– | ———– | ———– | ———– |
| Google | Bard, LaMDA, PaLM | Conversational + creative AI | Gradual launch |
| Microsoft | Prometheus Model | Bing search assistant | Limited testing after viral mishaps |
| Anthropic | Constitutional AI | Safe and helpful assistants | Private beta testing |
| OpenAI | GPT models | Creative content generation | ChatGPT widely available |
Many major tech companies are funneling money into this AI talent competition. 2023 is being called the “Year of Generative AI” as unprecedented progress gets demonstrated.
Whichever firm leads in advanced models like image and video generation may win billions in cloud revenue and set standards for the industry. Both tech giants and startups are racing to be king of this next AI paradigm shift.
The Road Ahead – Promise and Concerns
Bard has immense promise to enhance search, e-commerce, creativity and more. Its natural language and visual capabilities could aid everything from travel planning to medical research.
But like any transformative technology, AI comes with risks if deployed without enough care and debate. Issues around bias, misinformation and job loss will require ethical governance.
How Google manages these tensions as Bard develops will shape public trust. While excitement is high, skepticism remains on whether AI should automate certain human roles and tasks.
Ongoing progress will depend on transparency and open collaboration between companies, governments and civil society. With responsible development, Bard could spearhead an AI revolution powered by Google technology.
To err is human, but AI does it too. Whilst factual data is used in the production of these articles, the content is written entirely by AI. Double check any facts you intend to rely on with another source.
