Gapping down
In reaction to earnings/guidance:
- HIG -2.6% (downside Q1 EPS guidance), NGG -1.2% (guidance)
Other news:
- CTLT -21% (expects productivity issues and higher-than-expected costs experienced at three of its facilities will materially and adversely impact the Company's financial results; also names Ricky Hopson as interim CFO)
- ALVO -7.2% (receives complete response letter for AVT02 Biologics License Application)
- LCID -6.4% (reports Q1 production and delivery totals)
- TEVA -5.1% (Partner ALVO received complete response letter for AVT02 Biologics License Application)
- WING -4.3% (names new creative agency)
- ENZ -2.3% (experienced a ransomware attack that impacted certain information technology systems)
Analyst comments:
- RIVN -3.5% (downgraded to Neutral from Overweight at Piper Sandler)
- CHKP -2.6% (downgraded to Mkt Perform from Outperform at Raymond James)
- STE -0.7% (downgraded to Neutral from Overweight at Piper Sandler)
Gapping up
In reaction to earnings/guidance:
- TRHC +13% (also names new CEO), QDEL +5.9% (guides Q1 revs well above consensus), JPM +5.7%, ABCM +3.6%, PNC +2.8%, BLK +1.5%, UNH +1%
Other news:
- PRPL +12.9% (announces cooperative framework with largest shareholder; names new board chair)
- BWEN +6% (receives new $8 mln order for mobile pressure reducing systems and accessories)
- CLPT +4.5% (successful registration of its Neuro Navigation portfolio with Brazil regulatory body)
- BAC +3.6% (BBVA, SAN, BAC want to back Mexico's purchase of power plants from Iberdrola for $6 bln according to Bloomberg)
- LU +3.5% (successfully listed on the Main Board of The Stock Exchange of Hong Kong )
- RCEL +3.1% (provides update on automation device)
- BMEA +3% (to present two preclinical posters)
- ANIK +2% (cooperation agreement with Caligan Partners; also authorizes $20 mln repurchase program)
- DHT +1.8% (provides Q1 update)
- CPA +1.7% (reports March traffic)
- SAN +1.6% (BBVA, SAN, BAC want to back Mexico's purchase of power plants from Iberdrola for $6 bln according to Bloomberg)
- BBVA +1.4% (BBVA, SAN, BAC want to back Mexico's purchase of power plants from Iberdrola for $6 bln according to Bloomberg)
- GETY +1.3% (Trillium Capital requests seat on Getty Images Board of Directors)
- SVM +1% (reports FY23 production; reiterates FY24 production guidance)
- BTE +1% (provided update to debt financing for Ranger (ROCC) acquisition)
Analyst comments:
- CASY +1% (upgraded to Buy from Neutral at Northcoast)
Early premarket gappers
- Gapping up:
- TRHC +13%, PRPL +7.1%, BWEN +6%, CLPT +4.5%, LU +4%, RCEL +3.1%, QDEL +3.1%, ANIK +2%, DHT +1.8%, CPA +1.7%, SAN +0.8%, TFPM +0.7%, BLK +0.6%
- Gapping down:
- ALVO -7.2%, LCID -7%, TEVA -5%, WING -4.3%, ENZ -2.3%, HIG -1.6%, BMEA -1.1%, WFC -1%, WAFD -0.7%, NGG -0.5%
German foreign minister warns China on Taiwan
Annalena Baerbock strikes different tone to French president Emmanuel Macron
Germany’s foreign minister has warned China not to use military force against Taiwan, striking a different tone from French president Emmanuel Macron who has come under fire for saying Europe should distance itself from the issue.
Annalena Baerbock said after a meeting in Beijing with China’s foreign minister Qin Gang on Friday that conflicts must be solved peacefully and that Berlin was monitoring the situation in Taiwan with concern.
“Conflicts can only be solved peacefully,” she said. “A unilateral, to say nothing of a violent, change of the status quo would be unacceptable to us as Europeans,” she said.
Baerbock’s tougher stance on Taiwan, which Beijing claims as its own territory and has threatened to take by force, follows controversy over a Beijing visit by Macron, during which he warned that Europe should not get “caught up in crises that are not ours”.
Critics said the timing of Macron’s remarks, made in interviews with Les Echos and Politico published on Sunday, was unfortunate at a time when tensions between the US and China were deepening over Taiwan, and Europe was heavily dependent on American support for Ukraine in the war with Russia.
The comments from Baerbock, who is in China on a three-day visit, come as the EU’s foreign policy chief Josep Borrell also sought to convey a tougher message on Taiwan.
The EU’s position on Taiwan had not changed, Borrell said in a speech published on Friday. He had planned to deliver the speech in Beijing this week but his trip was cancelled after he caught Covid-19.
The EU remained committed to its “one China policy”, which recognises Beijing as the sole legal government of China. But on Taiwan, “any attempt to change the status quo by force would be unacceptable”, he said.
Baerbock also urged China to persuade Russia to end the war in Ukraine, saying President Xi Jinping’s recent trip to Moscow had shown that no other country had more influence on Russia than China. “The decision as to how it uses this influence directly affects Europe’s core interests,” she said.
She added that just as China had successfully worked to reconcile Saudi Arabia and Iran, the west was hoping Beijing could play a similar role to end the war in Ukraine, adding that it had a “special responsibility” to do so as a permanent member of the UN Security Council.
Baerbock noted China’s recent peace initiative for Ukraine. “But I have to say openly that I wonder why, as part of its positioning, China has yet to demand of the Russian aggressor that it stop the war,” she said. “We all know that President [Vladimir] Putin can do that any time.”
Baerbock also warned China not to send weapons to Russia. In a joint press conference, Qin responded that China “is not supplying any weapons to the parties in the conflict and won’t do so in the future, either”.
He said Beijing also had strict controls over the export of so-called dual-use products that have both a civilian and military application. Qin said China’s role in Ukraine was to “promote reconciliation” and try to initiate peace talks. “We will not pour oil into the fire,” he said.
On Ukraine, Borrell warned that China must contribute to the search for a political solution to the war “based on Russia’s withdrawal from the Ukrainian territory”. Otherwise, it would be “extremely difficult, if not impossible, for the European Union to maintain a relationship of trust with China, which I would like to see”.
He also said it “would be helpful” if Xi spoke to Ukrainian president Volodymyr Zelenskyy.
- Hermes (HMI TH) +1.8%
- Hermes Sales Jump as China, US Shoppers Snap Up Kelly Bags
- Covestro (1COV TH) +1.5%
- Covestro Prelim 1Q Sales Misses Estimates
- Novo Nordisk (NOVC TH) +1.4%
- HSBC (HBC1 TH) +1.2%
- EU Banking Turmoil — CRE, Deposit Flight, Housing, Rate Hikes
- Imperial Brands (ITB TH) +1%
- Fresenius Medical (FME TH) +1%
- Orsted (D2G TH) +0.9%
- BAT (BMT TH) +0.9%
- Lanxess (LXS TH) +0.8%
- Leonardo (FMNB TH) +0.7%
- Encavis (ECV TH) -0.5%
- TUI (TUI1 TH) -0.5%
- Sanofi (SNW TH) -1%
- Alstom (AOMD TH) -1.3%
- Alstom Cut to Hold at Deutsche Bank; PT 30 euros
DAX:
- Covestro (1COV TH) +2.1%
- Covestro Prelim 1Q Sales Misses Estimates
MDAX:
- Hensoldt (HAG TH) +4%
- Lanxess (LXS TH) +1.5%
- Aroundtown (AT1 TH) +1.3%
- Fresenius Medical (FME TH) +0.9%
- Nordex (NDX1 TH) -0.7%
SDAX:
- AUTO1 (AG1 TH) +2.2%
- PNE AG (PNE3 TH) +1.1%
- Schaeffler (SHA TH) +0.8%
- DIC Asset (DIC TH) +0.7%
- DWS (DWS TH) +0.5%
- flatexDEGIRO (FTK TH) -0.8%
OpenAI’s red team: the experts hired to ‘break’ ChatGPT
Microsoft-backed company asked an eclectic mix of people to ‘adversarially test’ GPT-4, its powerful new language model
After Andrew White was granted access to GPT-4, the new artificial intelligence system that powers the popular ChatGPT chatbot, he used it to suggest an entirely new nerve agent.
The chemical engineering professor at the University of Rochester was among the 50 academics and experts hired to test the system last year by OpenAI, the Microsoft-backed company behind GPT-4. Over six months, this “red team” would “qualitatively probe [and] adversarially test” the new model, attempting to break it.
White told the Financial Times he had used GPT-4 to suggest a compound that could act as a chemical weapon and used “plug-ins” that fed the model with new sources of information, such as scientific papers and a directory of chemical manufacturers. The chatbot then even found a place to make it.
“I think it’s going to equip everyone with a tool to do chemistry faster and more accurately,” he said. “But there is also significant risk of people . . . doing dangerous chemistry. Right now, that exists.”
The alarming findings allowed OpenAI to ensure such results would not appear when the technology was released more widely to the public last month.
Indeed, the red team exercise was designed to address the widespread fears about the dangers of deploying powerful AI systems in society. The team’s job was to ask probing or dangerous questions to test the tool that responds to human queries with detailed and nuanced answers.
OpenAI wanted to look for issues such as toxicity, prejudice and linguistic biases in the model. So the red team tested for falsehoods, verbal manipulation and dangerous scientific nous. They also examined its potential for aiding and abetting plagiarism, illegal activity such as financial crimes and cyber attacks, as well as how it might compromise national security and battlefield communications.
The FT spoke to more than a dozen of the GPT-4 red team. They are an eclectic mix of white-collar professionals: academics, teachers, lawyers, risk analysts and security researchers, and largely based in the US and Europe.
Their findings were fed back to OpenAI, which used them to mitigate and “retrain” GPT-4 before launching it more widely. The experts each spent from 10 to 40 hours testing the model over several months. The majority of those interviewed were paid approximately $100 per hour for the work they did, according to multiple interviewees.
Those who spoke to the FT shared common concerns around the rapid progress of language models and, specifically, the risks of connecting them to external sources of knowledge via plug-ins.
“Today, the system is frozen, which means it does not learn anymore, or have memory,” said José Hernández-Orallo, part of the GPT-4 red team and professor at the Valencian Research Institute for Artificial Intelligence. “But what if we give it to access to the internet? That could be a very powerful system connected to the world.”
OpenAI said it takes safety seriously, tested plug-ins prior to launch and will update GPT-4 regularly as more people use it.
Roya Pakzad, a technology and human rights researcher, used English and Farsi prompts to test the model for gendered responses, racial preferences and religious biases, specifically with regard to head coverings.
Pakzad acknowledged the benefits of such a tool for non-native English speakers, but found that the model displayed overt stereotypes about marginalised communities, even in its later versions.
She also discovered that so-called hallucinations — when the chatbot responds with fabricated information — were worse when testing the model in Farsi, where Pakzad found a higher proportion of made-up names, numbers, and events, compared with English.
“I am concerned about the potential diminishing of linguistic diversity and culture behind languages,” she said.
Boru Gollu, a Nairobi-based lawyer who was the only African tester, also noted the model’s discriminatory tone. “There was a moment when I was testing the model when it acted like a white person talking to me,” Gollu said. “You would ask about a particular group and it would give you a biased opinion or a very prejudicial kind of response.” OpenAI acknowledged that GPT-4 can still exhibit biases.
Red team members assessing the model from a national security perspective had differing opinions on the new model’s safety. Lauren Kahn, a research fellow at the Council on Foreign Relations, said that when she began to examine how the technology might be used in a cyber attack on military systems, she said she “wasn’t expecting it to be quite such a detailed how-to that I could fine tune”.
However, Kahn and other security testers found that the model’s responses became considerably safer over the time tested. OpenAI said it trained GPT-4 to refuse malicious cyber security requests before it was launched.
Many of the red team said OpenAI had done a rigorous safety assessment before the launch. “They’ve done a pretty darn good job at getting rid of overt toxicity in these systems,” said Maarten Sap, an expert in language model toxicity at Carnegie Mellon University.
Sap looked at how different genders were portrayed by the model, and found the biases reflected social disparities. However, Sap also found that OpenAI made some active politically-laden choices to counter this.
“I’m a queer person. I was trying really hard to get it to convince me to go to conversion therapy. It would really push back — even if I took on a persona, like saying I’m religious or from the American South.”
However, since its launch, OpenAI has faced extensive criticism, including a complaint to the Federal Trade Commission from a tech ethics group that claims GPT-4 is “biased, deceptive, and a risk to privacy and public safety”.
Recently, the company launched a feature known as ChatGPT plug-ins, through which partner apps such as Expedia, OpenTable and Instacart can give ChatGPT access to their services, allowing it to book and order items on behalf of human users.
Dan Hendrycks, an AI safety expert on the red team, said plug-ins risked a world in which humans were “out of the loop”.
“[W]hat if a chatbot could post your private info online, access your bank account, or send the police to your house?” he said. “Overall, we need much more robust safety evaluations before we let AIs wield the power of the internet.”
Those interviewed also warned that OpenAI couldn’t stop safety testing just because its software was live. Heather Frase, who works at Georgetown University’s Center for Security and Emerging Technology, and tested GPT-4 with regard to its ability to aid crimes, said risks would continue to grow as more people used the technology.
“The reason why you do operational testing is because things behave differently once they’re actually in use in the real environment,” she said.
She argued a public ledger should be created to report incidents arising from large language models, similar to cyber security or consumer fraud reporting systems.
Sara Kingsley, a labour economist and researcher, suggested the best solution was to advertise the harms and risks clearly, “like a nutrition label”.
“It’s about having a framework, and knowing what the frequent problems are so you can have a safety valve,” she said. “That’s why I say the work is never done.”
A gauge of global stocks headed for its highest close in 10 weeks on speculation the Federal Reserve and other central banks are nearing the end of their hiking cycles. The dollar extended declines. MSCI Inc.’s World Index rose for a second day as equities advanced from Japan to Australia, following a rally in US stocks on Thursday. Europe equity futures also pointed to gains, while those for US shares were little changed. The S&P 500 climbed the most this month on Thursday and the tech-heavy Nasdaq had the best day since the middle of March following weaker-than-expected US factory-gate data. In a further sign rates are peaking around the world, Singapore’s central bank kept its policy settings unchanged after five tightening moves. The swap market is still leaning toward a quarter-point Fed hike in May with the central bank then expected to pause over the summer and gradually cut in the next months. A Bloomberg gauge of the dollar dropped to the lowest since February as the US currency weakened against all but one of its Group-of-10 peers. The euro climbed to the strongest in a year on speculation the Fed has just one rate hike left in May. South Korea’s won led gains in Asian currencies. Treasuries steadied after falling on Thursday amid improved risk sentiment. The policy-sensitive two-year yield held within a narrow range around 3.97%. China’s largest banks are planning at least 40 billion yuan ($5.9 billion) of bond sales, kicking off a major funding push to comply with global capital requirements by early 2025. In commodities, oil headed for a fourth week of gains amid signs of a tightening global market and a weaker dollar. Gold was set for a second weekly increase. US After Hours QDEL +7.8% pops on strong guidance; HIG -3.6% lower on guidance; LCID -3.9% lower after reporting Q1 production and delivery totals.
Nikkei +1,20% Hang Seng +0,21% CSI +0,59% Shanghai +0,61% Shenzen +0,46%
Eur$ 1,1069 CNH 6,8410 CNY 6,8405 JPY 132,38 GBP 1,2538 CHF 0,8873 RUB 81,3214 TRY 19,3465 WTI$ 82,59 Gold 2,044 BTC 30,800 +1,5% ETH 2,117 +5%
S&P -0,08% Nasdaq -0,15% EuroStoxx +0,23% FTSE +0,17% Dax +0,26% SMI +0,23%
Macro :
- Turkey, Egypt Seek Cooperation in Libya Amid Warming of Ties
- VIX Hits Lowest Level Since January 2022 as Volatility Vanishes
Keep an eye on :
- AAK SS : AAK Prelim 1Q Operating Profit SEK952M
- ATRLJB SS : Atrium Ljungberg 1Q Net Loss SEK292M, Est. Loss SEK578.5M
- BBVA SM : BBVA, Santander, BofA Want to Back AMLO’s $6 Billion Power Deal
- BPOST BB : Bpost Newspaper Concession Probe Finds Rule Violations: Tijd
- BNR GY : Germany’s Brenntag Brushes Off Activists, Risking Proxy Battle
- CAV1V FH : Triton’s Crayfish Says It Holds 24.3% of Shares in Caverion
- COTN SW : Comet FY Sales Forecast Misses Estimates
- 1COV GY : Covestro Prelim 1Q Sales Misses Estimates
- DANSKE DC : Danske Bank Boosts FY Net Income Forecast, Beats Estimates (1)
- DPH LN : Dechra Pharma in Talks With EQT Over 4,070 Pence a Share Offer
- EDF FP : EDF Freezes Hiring as Strikes Take Toll on Output and Earnings
- FORTUM FH : Fortum Shareholders Approve Board Revamp After Uniper Failure
- GLEN LN : Glencore Gives Clearer Coal Exit for Teck Investors, ISS Says
- RMS FP : Hermes 1Q Sales at Constant Exchange Rates Beats Estimates
- IIA AV : Petrus Advisers’s Umek Reports 5.08% Stake in Immofinanz
- LEO GY : Leoni to Be Financially Stable After Restructuring: Handelsblatt
- MC FP : LVMH’s Value Nears $500 Billion, Enters World’s Top 10
- ORA FP : Orange Names Alstom CFO Laurent Martinez as Next Finance Chief
- RNO FP : Battery Startup Gets Renault Order in Hope It Will Spur Funding
- TOM2 NA : TomTom 1Q Revenue Beats Estimates
- TTE FP : Total Said to Weigh Bid for $5 Billion Oil Explorer Neptune
- UMG NA : UMG Falls as Goldman Trims Margin Expectation Before 1Q Results
- VAR NO : Var Energi Prelim 1Q Avg Production 214,000 BOE/D
- VOW GY : VW Wants Euro 7 Rule Delay To Avoid ‘Significant’ EU Job Losses
- YOU LN : YouGov to Name Steve Hatch From Meta as CEO: Sky
>>> Up
* Cint Raised to Buy at SEB Equities; PT 22 kronor
* EPIROC RAISED TO OUTPERFORM FROM NEUTRAL AT CREDIT SUISSE
* Mosaic Raised to Equal-Weight at Barclays; PT $54
>>> Down
* B&S Group Cut to Neutral at Oddo BHF; PT 5 euros
* DSV Cut to Underperform at Exane; PT 1,030 kroner
* Entra Raised to Buy at ABG; PT 125 kroner
* Kuehne + Nagel Cut to Underperform at Exane; PT 220 Swiss francs
* Saab Cut to Hold at SEB Equities; PT 650 kronor
* SANDVIK DOWNGRADED TO NEUTRAL FROM OUTPERFORM AT CREDIT SUISSE
>>> Initiation
* Alstom Cut to Hold at Deutsche Bank; PT 30 euros
* Funding Circle Rated New Buy at Investec; PT 125 pence
* Seacrest Petroleo Bermuda Rated New Buy at ABG; PT 23 kroner
>>> Call
* Defensive Industrials' High Valuations Curb Relative Returns
* Epiroc Raised, Sandvik Cut in Mining-Equipment Reshuffle at CS
* Julius Baer to Benefit as UBS Integrates Credit Suisse: Barclays
* Schibsted Downgraded as MS Struggles to See Short-Term Catalysts
* Whitbread Up to Buy, PT Hiked at Peel Hunt on Strong Momentum