AI Memmo | 11-19-23
Below is a summary of the most interesting articles from the past week related to AI.
article date: 11/19/2023
Microsoft Ignite 2023: Copilot AI expansions, custom chips and all the other announcements
Summary: Microsoft's Ignite 2023 event was highlighted by several AI product reveals and
expansions. The company
rebranded its AI-powered chatbot, Bing Chat, to Copilot, making it accessible in Windows and other
platforms. Microsoft
also announced its custom AI chips, Azure Maia 100 AI Accelerator and Azure Cobalt 100 CPU, aimed at
reducing GPU
dependency. New AI tools were unveiled, including Copilot for Azure, Copilot for Service, Copilot
Studio, and Copilot in
Dynamics 365 Guides. Microsoft Teams introduced an AI-powered home decorator and voice isolation
feature. The company
also consolidated its planning tools into a single product and announced copyright protections for more
customers using
Azure's OpenAI Service. Lastly, Microsoft introduced Windows AI Studio for running AI locally on Windows
and Azure AI
Speech for creating text-to-speech avatars.
Link to article
Category: AI in Business, Technology Updates.
Meta brings us a step closer to AI-generated movies
Summary: Meta has introduced a new technology, Emu Video, that brings us closer to
AI-generated movies. Emu Video is an
evolution of Meta's image generation tool, Emu, and can generate a four-second animated clip from a
caption, image, or
photo paired with a description. The clips can be edited using another AI model, Emu Edit, which allows
users to
describe modifications in natural language. Despite its impressive capabilities, Emu Video has
limitations, particularly
in animating complex actions and maintaining consistent object presence. The technology has raised
ethical concerns
about the potential displacement of animators and artists, as well as issues around copyright and
compensation for
creators whose works are used in training AI models.
Link: Link to Article
Category: Business, entertainment
Google is opening up its Bard AI chatbot to teenagers
Summary:
Google is expanding access to its conversational AI tool, Bard, to teenagers who meet the minimum
age requirement
to manage their own Google Account. The tool, which can be used for learning and problem-solving, will
be available in
English initially, with more languages to be added in the future. Google has implemented safety features
and guardrails
to protect teen users from inappropriate content. Bard can also assist with data visualization and
provide step-by-step
explanations for solving math equations. The first time a teen asks a fact-based question, Google's
double-check
response feature will automatically run to evaluate the accuracy of Bard's response. The company is also
planning to
introduce an onboarding experience for teens to learn how to use generative AI
responsibly.
Link: Link to Article
Category: Generative AI
Siena AI raises $4.7M to develop an empathic AI customer service agent
Summary: Siena AI, a company co-founded by Andrei Negrau and Lisa Popovici, has raised $4.7
million in
seed funding to
develop an AI customer service agent that can respond with empathy like a human. The AI-powered customer
support
solution is designed to understand context and juggle multiple tasks in the same interaction. The
company's unique
features include AI Personas, which allow merchants to create a unique brand voice, and a cognitive
reasoning-based
engine, CORE, which determines the best resolution path for complex customer service problems. Since its
launch six
months ago, Siena AI has acquired 65 customers and manages up to 80% of customer interactions across
over 100 languages
and all channels. The funding will be used for hiring, go-to-market capabilities, and technology
development.
Link:
Link to Article
Category: Business
Microsoft launches a deepfakes creator at Ignite 2023 event
Summary: Microsoft has launched a tool called Azure AI Speech text-to-speech avatar at the
Ignite 2023
event. This tool
can create a photorealistic avatar of a person and animate it to say things that the person didn't
necessarily say.
Users can generate videos of an avatar speaking by uploading images of a person and writing a script.
The tool trains a
model to animate the avatar, while a separate text-to-speech model reads the script aloud. The avatars
can speak in
multiple languages and can be used to build conversational agents, virtual assistants, chatbots, and
more. Microsoft has
implemented measures to prevent misuse, including limiting access to custom avatars and requiring
explicit written
permission from avatar talent.
Link:
Link
to Article
Category: Business
Google claims it is beating Huawei in AI weather prediction
Summary: Google's DeepMind AI unit has developed an AI weather forecaster, GraphCast, that
has
outperformed the European
Centre for Medium-Range Weather Forecasting (ECMWF) across 90% of over 1,300 atmospheric variables such
as humidity and
temperature. Unlike traditional weather simulations that require powerful servers and hours to compute,
GraphCast can
generate a forecast in under a minute on a laptop. The AI model uses machine learning algorithms trained
on 39 years of
observations collected by the ECMWF. Despite its success, the model has limitations, including a
tendency to
underestimate the strength of significant events like Category 5 storms and difficulties in predicting
conditions in the
stratosphere.
Link:
Link
to Article
Category: Business
Biden’s Executive Order on AI Is a Good Start, Experts Say, but Not Enough
Summary: President Joe Biden has signed an executive order that sets new federal standards
for AI
safety, security, and
trustworthiness. The order, which is the most comprehensive official policy on artificial intelligence
in the U.S. to
date, also addresses various aspects of AI risk and development. It mandates AI developers to share
safety data,
training information, and reports with the U.S. government before publicly releasing large AI models or
updated versions
of such models. The order also requires the creation of federal standards and tests to ensure AI doesn't
threaten
national security. However, experts argue that while the order is a significant step forward, it has
notable gaps and is
insufficient to tackle all the problems posed by advancing AI.
Link:
Link
to Article
Category: Business
US, China at critical odds on future of military AI
Summary: The US and China have agreed to hold talks on the future use of artificial
intelligence (AI) in
military areas
amid rising tensions over various issues. The agreement was made during the Asia Pacific Economic
Cooperation (APEC)
summit. However, there is no indication that the two countries are close to making a joint declaration.
The discussions
are expected to focus on the risks associated with using advanced AI systems in sensitive military
functions. Despite
this, neither side is likely to accept any ban limiting their freedom to deploy AI in their militaries.
The US is
pushing for clear and strong legal instruments that restrict the use of AI in autonomous weapons
systems, while China
aims to counterbalance and rival the US in setting future AI standards, particularly in the military
sphere.
Link:
Link
to Article
Category: AI in Military, International Relations.
Mind-Reading AI Turns Thoughts Into Text, Gives ALS Patient His Voice Back
Summary: Unbabel, a neural tech company, has demonstrated its Project Halo at the Web Summit
conference.
The technology
aims to enable silent, thought-based communication between humans and machines by combining a
non-invasive neural
interface with generative AI to transform bioelectrical signals into language. The system allows users
to receive a
message via earbuds and then send a reply by simply thinking about what they want to say. This
technology has potential
applications ranging from replying to texts in a movie theater to helping people with amyotrophic
lateral sclerosis
(ALS) communicate via text or audio notes. The company plans to commercially launch Project Halo in
2024.
Link:
Link
to Article
Category:
AI in Healthcare, Neural Interfaces.
ChatGPT Puts Paid Plan Signups on Pause as Systems Strain
Summary: OpenAI has temporarily halted new signups for its ChatGPT Plus program due to
overwhelming
demand that has
exceeded the platform's capacity. The decision was announced by the company's CEO, Sam Altman, following
a surge in
consumer demand after the introduction of new features at the company's developer conference, DevDay.
The company also
shared details about its new GPT-4 Turbo AI model at the event and discussed plans to pay creators who
are subscribed to
the program for using GPTs they have made. The pause in new signups is a response to the need to ensure
a high-quality
user experience amidst the surge in usage.
Link:
Link
to Article
Category: AI Services, OpenAI.
New Open Source AI Model from China Boasts Twice the Capacity of ChatGPT
Summary: Beijing Lingyi Wanwu Information Technology Company has developed an open-source
artificial
intelligence (AI)
model, known as the Yi series, which can handle up to 200,000 tokens of context. This capacity
significantly surpasses
other popular models like Anthropic's Claude and OpenAI's GPT-4 Turbo. The Yi series, which includes the
lightweight
Yi-6B-200K and the more robust Yi-34B-200K, can understand and respond to English and Mandarin. Despite
a limitation
where the model struggles to retrieve accurate information when a prompt occupies more than 65% of its
capacity, the Yi
series performs well in reading comprehension, common-sense reasoning, and common AI tests. The
open-source nature of
the model allows users to run it locally on their own systems, offering greater control and
customizability.
Link:
Link
to Article
Category: AI Development, International AI.
Send comments and feedback here.
<<< Blog Homepage