2024-07-26 02:35:46
KI-Ticker
Google Gemini introduces far-reaching improvements
Updated on 25.07.2024 – 18:00Reading time: 8 min.
New reports and updates on AI appear every day. We always summarize the most important developments for you.
Artificial intelligence (AI) is developing rapidly and changing our everyday lives. From ChatGPT to AI-controlled self-driving cars – the technology is making enormous progress. We have summarized the most exciting innovations and most important recent developments for you.
Google is bringing new features to its AI assistant Gemini. Users can now access the faster version 1.5 Flash for free in over 40 languages and 230 countries. According to Google, this version is particularly better at argumentation and image understanding. The context window has also been enlarged so that longer conversations will be possible in the future.
Another new feature is the display of related content in Gemini’s answers. When the assistant provides facts on a topic, links to further information are now also displayed. This feature is intended to help reduce misinformation and make research easier. The double-check function, which checks answers using Google search, is also being expanded.
Google also says it is expanding access to Gemini for teens ages 16 and up in more than 40 languages. The company emphasizes that the safety and developmental needs of young people are a priority and that additional guidelines and safeguards have been put in place.
Meta has released its new AI language model Llama 3.1 as open source. Company boss Mark Zuckerberg calls it a “smash hit” and says it can keep up with GPT-4o, but is cheaper. Meta wants to become a pioneer and sees the future of AI development in the open provision of the software.
Zuckerberg explains the benefits of open source in a blog post: It promotes innovation and independence. He also sharply criticizes Apple: “With the way they tax developers, the arbitrary rules they apply, and all the product innovations they block, it’s clear that Meta and many other companies could build much better services for people.”
Meta also sees opportunities for greater security in disclosure. Zuckerberg distinguishes between unintended risks such as incorrect AI advice and intentional misuse. He believes that transparency and broad monitoring could help minimize these risks. Meta also plans to provide developers with tools for their own AI agents.
According to a recent analysis by Europol, millions of people fall victim to online crime every day. The experts at the EU police authority warn in particular about cyberattacks, sexual exploitation of children and financial fraud. It is also worrying that more and more young offenders are appearing.
Small and medium-sized companies in particular are the focus of hackers because their digital security measures are often less robust. But private individuals are not spared either: phishing attacks, dating scams and fake online shops make life difficult for many. According to Europol, the increase in sexual abuse of children on the Internet is particularly alarming.
Investigators are also concerned about the use of artificial intelligence by criminals. The technology is apparently being used for increasingly sophisticated blackmail methods. Gangs even sell ready-made templates for fraud or sexual exploitation online. So-called deepfakes, deceptively real fakes of images and sounds, are also being used more and more – for example, to blackmail relatives with fake emergency calls or to manipulate images of child abuse.
The German start-up DeepL has introduced a new AI model for translations. According to the company, the system delivers more accurate results than the competition from Google, Microsoft and OpenAI. The Cologne-based company refers to blind tests by language experts who are said to have preferred DeepL translations more often.
Unlike other AI models, DeepL’s new technology was specifically developed for translation. The company says it uses special data collected over seven years. In addition, “thousands of hand-picked and specially trained language experts” were involved in the optimization.
The new model is initially only available to DeepL Pro customers for English, Japanese, German and simplified Chinese – more are to follow soon. DeepL offers paid solutions for companies as well as a free version for short texts. The start-up employs around 900 people and is valued at around 1.85 billion euros.
More and more companies in Germany are using artificial intelligence in their workflows and business processes. According to the Munich-based Ifo Institute, citing the results of its economic surveys, 27 percent of companies currently use AI – a year ago, the figure was less than half (13.3 percent).