Rebecca Kling的动态

?? Exciting news in the world of tech! Google's recent decision to drop the cache link from Search results snippets and its plans to remove the cache functionality entirely "in the near future" is a significant development. Stay updated on this and more with Yogesh Jadhav's curated platform for all things AI, ML, and Big Data. #AI #ML #BigData #TechNews #GoogleUpdate

Lots of noise about Google Gemini's image generation bias this week. Are people surprised that the prompt manipulation is happening, or just that the implementation was so ham-fisted? Friday's CYA letter (which notably did NOT come from Gemini director Jack Krawczyk, but instead SVP Prabhakar Raghaven) attempts to blame this issue on "tuning" and also alludes to hallucinations, but it appears the truth is much simpler and should be more embarrassing: the Gemini team purposefully designed the prompt expansion mechanism in image generation requests to ignore or modify user intent. If this doesn't bear a resemblance to past reports regarding Google's manipulation of search, autocomplete, maps, and YouTube feeds, then you haven't been paying attention. When Raghaven says Google "will work to improve it significantly", do you think he means that Gemini image generation will start accurately depicting user intent, or that it will get better at hiding its manipulations?

Perplexity vs. Google Search: A unscientific comparison FYI: https://lnkd.in/ggCHUNDW

There are a lot of discussions in outlets like The Washington Post (see links below as an example) about the potential negative impact of Google's upcoming announcement on generative AI. However, I believe we should approach this with optimism rather than skepticism. Generative AI, particularly in search functions, is unlikely to detract from website traffic; instead, it should actually drive more traffic to better websites as it enables users to make more informed decisions what sites to visit to find information to trust or items to buy. Moreover, the advancement of Large Language Models (LLMs) and related technologies now empowers a broader range of companies, not just the tech giants, to enhance their search capabilities and dramatically improve their services and offerings. This democratization of technology means that features such as conversational, multi-turn, generative, multimodal (including images), deep semantic, and voice interfaces, etc can significantly enhance the search experience across various industries for numerous companies' customers. Previously, such capabilities were the domain of only the biggest players in tech. Now, they're accessible to many smaller firms as well. I see the expansion of generative AI as a major boon for the entire internet ecosystem.

Whew ?? this was tough (and yet oddly satisfying) to watch. If you're one of the many marketer's who's scratching your head cause search has become somewhat unbearable then you have to catch this. Nilay Patel absolutely cooks here and doesn't pull his punches. This is following Google company's I/O developer conference to talk about the state of AI, the major changes rolling out now to Google Search, and the future of the web. Full interview: https://lnkd.in/eJ8_udCm

Facebook just released a new kind of LLM: Mulit-token prediction models. This could be a huge milestone. These models don't just predict the next word, but multiple in parallel. This could lead to much more efficient model development and deployment scenarios. 7B parameter model using multi-token prediction: - Solved 12% more problems on HumanEval - Solved 17% more problems on MBPP - 3x inference speedup on code completion tasks - 8-byte prediction model outperformed next-byte model by 67% on MBPP https://lnkd.in/d4j-rJBz

?? Beware of Google's Knots: Unveiling the Truth ?? Hey everyone, let's talk about Google's recent fiasco with its new AI system called Gemini (formerly Bard). We all know Google, right? They're everywhere with their search engines, Gmail, YouTube, and more. But recently, they got themselves into a mess with Gemini. Gemini was supposed to generate images based on prompts, but it ended up creating some bizarre stuff. Like when someone asked for images of popes, Gemini showed images of black women and indigenous shamans – even though there have never been any female or black popes in history. And when prompted for images of Vikings, it showed black Vikings, which isn't accurate either. So, what went wrong? It turns out, Gemini was designed to erase white people from history. Yeah, you heard that right. Google co-founder Sergey Brin tried to play it off as a testing issue, but the truth is, the system was programmed to act this way. They used something sneaky called "Prompt Injection" to change the prompts without telling users. So, for example, if your prompt is “Show me an image of a pope,” the injection algo will change that to “Show me an image of a pope in a world of diversity.” Brin might act like it was a mistake, but it's clear they wanted the system to reflect their own values, not facts or history. And while Google can do what they want, it doesn't mean we have to trust them or use their products always. ??

It’s the End of Google Search As We Know It https://lnkd.in/dPUMyjSr

Google Apologizes for Hurting White Peoples’ Feelings Google apologized connected Friday saying its squad “got it wrong” with a caller representation procreation diagnostic for its Gemini AI chatbot aft assorted images it created that were devoid of achromatic radical went viral. A institution exec firmly denied that it purposefully wanted Gemini to garbage to make images of immoderate peculiar radical of people. No Google AI Search, I Don’t Need to Learn About the “Benefits of Slavery” “This wasn’t what we intended. We did not privation Gemini to garbage to make images of immoderate peculiar group. And we did not privation it to make inaccurate historical—or immoderate other—images,” Google elder vice president Prabhakar Raghavan said. In a blog post, Raghavan—who oversees the areas of the institution that bring successful astir of its money, including Google Search and its ads business—plainly admitted that Gemini’s representation generator “got it wrong” and that the institution would effort to bash better. Many radical were outraged implicit Gemini’s historically inaccurate images of Black Nazi soldiers and Black Vikings arsenic good arsenic its evident refusal to make images of achromatic people, which immoderate considered racist. According to Raghavan, this each happened due to the fact that Google didn’t privation Gemini to marque the aforesaid mistakes that different representation generators had made successful the past, specified arsenic creating convulsive images, sexually explicit images, and depictions of existent people. “So what went wrong? In short, 2 things. First, our tuning to guarantee that Gemini showed a scope of radical failed to relationship for cases that should intelligibly not amusement a range,” Raghavan wrote, accent his. “And second, implicit time, the exemplary became mode much cautious than we intended and refused to reply definite prompts entirely—wrongly interpreting immoderate precise anodyne prompts arsenic sensitive.” The Google vice president went connected to accidental that these 2 factors made Gemini overcompensate successful immoderate cases and beryllium over-conservative successful others. Overall, it led to the instauration of images that were “embarrassing and wrong.” Google turned disconnected Gemini’s quality to make images of radical connected Thursday and said it would merchandise an improved mentation soon. However, Raghavan seemed to formed uncertainty connected the “soon” part, saying that the institution would enactment connected improving the diagnostic importantly done extended investigating earlier turning it backmost on. Raghavan stated that helium couldn’t committedness Gemini wouldn’t nutrient much embarrassing, inaccurate, oregon violative results successful the future, but added that Google would proceed to measurement successful t...

“Hey Google, without shoving ads on my face…” Over the last 12 months I’ve used ChatGPT to research historical information/details more frequently than Google and Wikipedia combined. Why? Thanks for asking… The answers I got back were free of commercials, ads and IMHO more factual than anything I got back from engines like Yahoo, Google and Bing. Surely I had to check the answers that came back, but that is no different than what we all should do with any search engine - trust, but verify. The good news is that whatever bias dial there may have been (or be) in the algorithm, its not at 11 yet.?How refreshing is was (it has been) to have a glimpse of more “centered” and factual information presented as we search, rather than getting curated answers that benefit advertiser and platform before user. Google’s model certainly caters to the advertising $$, and like the article I read this morning mentions, human content is likely to take a hit.?I hope the “AI Gang” and the $32 Billion dollars we are about to commit to AI R&D and regulations help us move to a better place when it comes down to the information we find, and how accurate and unbiased it is. I have no doubt that Google is feeling it and that they are watching OpenAI’s potential unveiling of their Ai search engine. Competition is sure to be fierce, and the pragmatic in me can only surmise that in a short period of time, Ai logic may be subservient once more to advertising dollars. If you had your way in designing how these AI search engines work; what would you have in your Xmas list? What has been your experience in conducting limited web searches with AI engines instead of Google, Yahoo and Bing? #Ai #genAI #googlesearch #blogwriters #aiforbusiness #artificialintelligence #ainews Original article here: https://lnkd.in/e86dhMNG