Bard enters the chatbot race…kinda

AI Blog

Issue #18
19/07/2023

Bard is now available in Europe! 

Bard, Google’s AI chatbot, is now available in Europe. We will be honest with you. It’s not the best thing out there. Especially when you factor in the plethora of AI chatbots that now exist and the funding that Google has at its disposal. However Bard has 2 very key advantages:

  1. Has access to the internet
  2. It is multimodal, i.e. it accepts images as input. Which gives us some very interesting use cases:
    • Sorting scanned invoices → here an example of how it was used as an expense management assistant
    • This is a great way to evaluate images of products or real estate
    • Documenting meeting notes
    • Transcribing text from real books

Since we now also have access to Bard, we find it useful to do a comparison of the available AI chatbots.

The table below shows us the features and what each AI chatbot can do:

 

  • GPT-4 → writing, complex & creative tasks
  • Claude-2 → analysis of docs/PDFs
  • Bing Chat → research requiring real sources
  • Bard → experimentation

 

Behind the scenes of ΑΙ

As we know, in order to improve the outputs of AI models, the reinforcement learning method is utilized. In essence, people evaluate the outputs as good or bad. That is why AI companies have hired professional raters. However, recently, the evils of these processes have surfaced. Google raters state that they are incredibly overworked and confused, as they are imperatively asked to rate content that they are not pre-qualified for, such as laws and medical advice. They are asked to flag false, deceiving, or misleading content, yet they are not given the proper resources to identify what that content is. OpenAI in turn, outsource content moderation to Kenya through Sama. In fact the raters are appealing to the country’s legislatures to launch investigations into Big Tech companies that outsource content moderation and AI work in Kenya. They say due to the nature of their work they are forced to read highly graphic and toxic content including scenes of murder, bestiality and rape, yet they are not provided with any psychological support. We feel it is important to know all aspects of the development of the AI models we use, so that we can be critical of them and not let the hype carry us away.

 

Resources

AI Insider 📰

  1. He said it and did it. Elon Musk pitched xAI, his own AI company with the goal of “understanding the true nature of the universe”.
    • The xAI team is filled with engineers/researchers from OpenAI and Google.
    • Because of Elon, xAI has access to the vast amount of text, images and videos from Twitter and Tesla to use as training data.
  2. Surprise, surprise Meta has published a research paper on yet another AI tool. The new CM3Leon (pronounced “chameleon”) is an image-generation AI model that actually has all the features and capabilities of the most popular image gen AIs. It can be:
    • generate images→ Meta even claims that it does it faster than anyone else
    • edit photos → text prompts, e.g. alter the color
    • place objects → needs a lot of precision
    • create captions & answer questions about the images
  3. Microsoft  has announced Bing Chat Enterprise and pricing for Microsoft 365 Copilot, which will cost $30 per user per month for Microsoft 365 E3, E5, Business Standard and Business Premium customers.
  4. OpenAI struck partnerships with Shutterstock and the The Associated Press. It will provide them with “priority access” to its technology in return for access to their files to use as training data.
  5. AI startups have already begun staff reductions
  6. Designers are suing Shein for AI rip-offs of their projects.
  7. Hollywood actors are also joining the writers’ strike, as Hollywood studios say they want to have their AI replicas forever and for free – Joan said what; here’s the speech by Fran Drescher, president of the actors’ union. She characteristically states: “You cannot exist without us!” And she’s not wrong, since without actors and writers, productions are frozen forever. Fun fact: The two groups haven’t held a strike together in 63 years.
  8. Over 1 billion images have been created with Adobe Firefly, which now supports prompts in over 100 languages. 
  9. Shopify announced Sidekick, the new AI assistant for entrepreneurs.

Learning Bytes 🧐

  1. Wipro, an Indian outsourcing company, is pouring $1 billion into training its staff in AI technologies.

  2. Schools in Hong Kong are teaching how to use AI tools and stand critically by their results.

  3. Research has found that writers complete their tasks, such as press releases, 40% faster and 18% better when using ChatGPT.

  4. FTC launched an investigation into OpenAI, into whether ChatGPT has harmed people through the spread of false information.

Cool Finds 🤯

Extras

  1. Why do AI detectors believe that the US Constitution is written by AI?
  2. A brief explanation of what the Coalition for Content Provenance and Authenticity (C2PA) is.

 

See you next week! 💚

#TogetherWeAI

google-PremierPartner-2024
facebook-pic
workplace-pic
agency-pic
social-media-pic

51

Social Media Awards

evo-pic

226

E-volution Awards

peak-pic

29

Peak Performance Marketing Awards

dime-pic

1

Effie Awards