DAI#42 – Hot chips, bad memory, and AI ups and downs
Welcome to this week’s roundup of triple-distilled AI news. This week, battle lines were drawn in the chip wars in Taiwan. Microsoft may need to forget to remember. And AI techs are fighting for their right to warn us about what they saw. Let’s dig in. Made in Taiwan This week all eyes were on the Computex 2024 event in Taiwan, where most of the world’s high-end AI chips are produced. GPU top dog NVIDIA unveiled its roadmap for increasingly powerful GPUs and other next-gen AI tech. CEO Jensen Huang says the company will be releasing a new family of The post DAI#42 – Hot chips, bad memory, and AI ups and downs appeared first on DailyAI.
Welcome to this week’s roundup of triple-distilled AI news.
This week, battle lines were drawn in the chip wars in Taiwan.
Microsoft may need to forget to remember.
And AI techs are fighting for their right to warn us about what they saw.
Let’s dig in.
Made in Taiwan
This week all eyes were on the Computex 2024 event in Taiwan, where most of the world’s high-end AI chips are produced.
GPU top dog NVIDIA unveiled its roadmap for increasingly powerful GPUs and other next-gen AI tech.
CEO Jensen Huang says the company will be releasing a new family of GPUs every year. The Blackwell GPUs are barely shipping yet, but Huang said an upgraded version will launch later this year.
When you take a closer look at the Blackwell performance numbers, you begin to get an idea of the potential impact NVIDIA’s new AI superchip will likely have.
AMD is still playing catchup in the AI chip race but also announced that it will release new generations of AI GPUs annually. With its sights set on NVIDIA, AMD says its new AI chip, the Instinct MI325X GPU beats NVIDIA’s H200 flagship GPU.
The specs are seriously impressive, but it must have been a little awkward bragging about beating the H200 when the Blackwell chips are almost ready to ship.
The AI chip battle between NVIDIA and pretty much every other computing company ratcheted up a notch last week. Big Tech companies, including AMD and Intel, formed a consortium to develop an open AI connectivity standard.
Guess who wasn’t invited to participate?
Forget about it
The Silicon Valley mantra of ‘Move fast and break things’ was in evidence this week.
Microsoft is about to unleash its Windows AI Recall feature which sounds great on paper.
Recall remembers everything you’ve seen and done on your PC and can help out when you can’t remember on which website you saw that great deal, or where you saved your lolcat pics.
But there’s a problem. Tech security experts are saying the way Recall stores the data could be a security disaster. It’s worse than you think, and Microsoft may want to recall Recall.
Microsoft’s enterprise customer briefings about Recall pic.twitter.com/q7eqnHKRwN
— Kevin Beaumont (@GossiTheDog) June 5, 2024
Last week Google’s AI Overviews told users to put glue on pizza and suggested eating rocks. Google finally admitted AI Overviews has a few snags but says it’s mostly your fault.
If Big Tech can’t get AI search and operating systems right, how will they tackle humanity’s big problems?
At the AI For Good Summit, industry leaders discussed examples of AI’s societal benefits and the challenges that need to be overcome to avoid widening the digital divide.
Will AI solve some of humanity’s biggest problems or make them worse?
OpenAI’s ups and downs
What happens when we become completely dependent on AI tools only for them to break? This week we got a taste of what that could be like.
OpenAI’s ChatGPT had a major outage as users couldn’t access the chat platform for hours. Was it a coincidence that Gemini, Claude, and Perplexity also had issues around the same time?
Could it have anything to do with supporters of disgruntled employees or AI skeptics?
Former and current OpenAI and Google employees just published a letter demanding the right to warn about AI risks without consequences.
The “Right to Warn” letter makes some good points but I’m not sure how likely it is that OpenAI and others will commit to their demands.
Fake politics
Lying and politics go together like fire and smoke. There never seems to be one without the other.
AI has supercharged people’s ability to spread disinformation but has also provided a handy excuse for politicians in a tight spot.
Conservatives are demanding potentially embarrassing audio of Joe Biden’s special counsel interview. The Justice Department says it won’t release the audio because it could be manipulated by AI.
Fair enough, or grabbing at AI straws?
AI Events
If you’re looking to be at the forefront of what’s happening in AI then you’re spoilt for choice with events like these happening around the world.
If you want to see how AI can give your marketing efforts a boost then check out the AI for B2B Marketers Summit 2024.
When you think about AI and electric vehicles you immediately think of Tesla and self-driving. But AI also helps with thermal management in EVs. The European EV Thermal Management Summit 2024 explores the emerging trends and tech in this space.
The AI_dev: Open Source GenAI & ML Summit Europe is a dynamic two-day event focusing on open-source generative AI and machine learning. We’ve got a great discount code for you if you’d like to attend.
In other news…
Here are some other clickworthy AI stories we enjoyed this week:
- LLMs have favorite numbers, and they’re not as random as you’d expect.
- Google DeepMind, Anthropic, and Microsoft AI attend the secretive Bilderberg Meeting.
- Avital Balwit, the chief of staff at Anthropic says 2024 is the first of the last five years she expects to work.
- Elon Musk ordered Nvidia to ship thousands of AI chips reserved for Tesla to X and xAI.
- OpenAI bans GPT accounts as it removes influence operations tied to Russia, China, and Israel.
- Nvidia hits $3 trillion market cap on back of AI boom.
And that’s a wrap.
At the rate new generations of GPUs are being released, 2025 is set to deliver a huge leap in AI advances. Can you imagine looking back a year from now and thinking how quaint GPT-4 is compared to GPT-5 and GPT-6?
Will you be using Windows AI Recall when it rolls out? I might, but only once they stop storing the data in plain text in an unsecured folder. What were they thinking?
The “Right to Warn” letter makes for chilling reading. When the people who built the AI say they’ve seen things they’re not allowed to warn us about, we should probably pay attention.
But Yann LeCun says there’s nothing to worry about so maybe it’ll be just fine.
Is it time to panic, or time to cheer on the AGI and chill? Let us know what you think, and send us links to any interesting AI news stories we missed.
The post DAI#42 – Hot chips, bad memory, and AI ups and downs appeared first on DailyAI.